How to Build & Sell AI Automations: Ultimate Beginner’s Guide

Transcript

00:00

In a world being transformed by AI, one skill stands above all others. AI automation.

Master this and you won't just survive the AI revolution, you'll thrive in it. I'm living proof of this.

Just 2 years ago, I taught myself how to build no code AI automations without any prior experience. And since then, I've

00:16

built multiple AI businesses, generated millions of dollars in revenue, and grown this channel to over 500,000 subscribers, and built AI systems for some of the biggest brands in the world. It's pretty safe to say that learning how to build AI automations has completely changed my life.

So, in this full course, I'll teach you everything

00:32

that I've learned about building AI automations and making money with them, even if you don't know how to code. And as they say, AI will not replace you, but the person using AI will.

So, my hope is that with this video, you too can learn this incredibly powerful skill to build the life of your dreams before it's too late. And as you can tell by

00:48

the length of this video, I'm not going to be holding anything back. So, I've split it into three different chapters.

Firstly, we'll build your foundational understanding of AI automation, covering what it actually is, the different types of AI automations, how they work under the hood, and the key concepts you need to know before we start building.

01:04

There's no technical background required to understand any of what I'm going to teach you there. Secondly, we'll dive deep into building out actual AI automations, taking you over my shoulder every step of the way as we build some of the most in- demand AI automation use cases in the market today.

This includes building things like cuttingedge voice

01:19

agents, too. And in the third and final chapter, I'll be giving you my proven blueprint for monetizing your AI automation skills while this technology explodes.

I'll share the exact strategies that I've used to generate millions of dollars with the skill set. So, if you're new to the channel and don't know who I am, let me quickly share why I am qualified to teach you

01:35

about AI automations in the first place. So, my name is Liam Mley and just 2 years ago, I started learning AI with no prior experience in the field.

Teaching myself how to build AI automations and chatbots through my own self-study, which I documented here on this YouTube channel from day one. This led me to starting Morningside AI, my AI

01:51

automation agency, where we build AI systems and agents for businesses from basic customer support systems when we started to now full AI SAS platforms for some of the biggest brands in the world. And I also have my own AI SAS called Agentive, which has over 70,000 users on it.

At Morningside AI, we've worked with publicly traded companies and even an

02:06

MBA team recently. And I also run the world's largest AI automation and business community with over 180,000 members on school.

So through this community and my YouTube channel, I've taught hundreds of thousands of people from all backgrounds how to build and make money from AI automation. And everything I'm about to teach you today

02:22

is exactly what helped me to achieve all of this. So let's dive in.

So there's a lot to cover here. I don't want you to give up halfway.

So let's quickly get clear on why learning AI automation is one of the most valuable skills anyone can have over the coming decade. Whether you're a student, an employee, or an entrepreneur.

Here's some quick truths

02:38

about AI and jobs. McKenzie predicts that AI and automation can replace up to 50% of current work activities by 2030.

And the World Economic Forum states that 41% of companies plan to reduce staff due to AI. Now, this is a lot of doom and gloom and many are naturally worried about their career in the future when

02:54

they hear the stuff, but it's not actually all bad if you know where to look. So, on the flip side of this same data, these same reports reveal an enormous opportunity for those willing to seize it.

The board economic forum's future of job report states that 50% of employees plan to reorient their business in response to artificial

03:09

intelligence and 66% of employees plan to hire talent with specific AI skills such as AI workflow automation. So on one hand we have the expectation of massive layoffs and automation of work over the next 5 to 10 years.

But on the other we have the majority of employers searching for people who have AI skills

03:26

or really just some form of basic AI literacy. Why is this?

Well, it's because AI literate individuals who can identify opportunities for automation and automate them themselves can have 5 to 10x the output of someone who doesn't know this and can't automate their own work. And I promise you that brushing up on your AI and actually becoming AI

03:42

literate so that you can be on the winning side of this next 5 to 10 years is so much easier than you think. I mean, it's it's literally as easy as watching this entire video in order to build your AI skills base.

If you don't believe me when I say that a little bit of self-study like this video goes a long way, here is an excellent clip from

03:57

the All-In podcast from one of the most respected investors and technologists in the world, Naval Ravakant, alongside a whole bunch of other biners. Again, I would say the easiest way to see that AI is not taking jobs or creating opportunities is go brush up on your AI, learn a little bit, watch a few videos,

04:14

use the AI, tinker with it, and then go reapply for that job that rejected you and watch how they pull you in. This video is exactly what Naval is talking about.

This is why I create these. So whether you're a student wanting to stand out in a competitive job market or an employee aiming to become irreplaceable at work or an entrepreneur

04:30

like me looking to scale your business with cuttingedge tools and automations, I have made this video for you. Now close out of all your other tabs, get a notebook and a pen and a beverage of your choice and make a commitment right now to yourself to finish this training and to ensure that you're going to be empowered by AI and not replaced by it.

That is all I want out of this video for

04:46

you all. So if you've done all that, let's get stuck into it.

All right. So, step one in building AI automations is knowing what an automation actually is.

And the term gets thrown around a lot these days. First thing we need to realize is that the AI part of the term is is relatively

05:01

new. Automation itself has been around for a long time.

So, let's start there to make this super easy to grasp. In simple terms, an automation is a system that does a task for you without you having to lift a finger.

It's kind of like setting up a little robot to do boring, repetitive stuff automatically, so you don't have to waste your time on

05:16

it. These are what we'll call old school automation.

The kind that existed way before chat GBT came along. They were often built on platforms like Zappia and lots of small to medium-sized businesses used them for the past 5 10 years.

And they did basic things like automatically saving info, for example, when someone filled out a form on a website. You

05:32

could make an automation that would take their name and their email and then just pop it into a spreadsheet. So just a little automation automating that boring stuff.

Or for example, when an email came in, it would send a quick alert to a chat app like Slack. So it's kind of like having a little helper who's following a super simple checklist.

If this happens, then do that. No thinking,

05:48

just just doing, right? And the benefit of this is huge.

It freed humans from doing super basic and boring work. And for business owners, it meant not having to pay more people just to handle these tiny and annoying tasks.

It's kind of like having a tireless assistant who never complains about doing the same boring thing over and over and over

06:04

again. So for decades, all was well in the automation space.

And these old school automation saved time and money, and everyone was super happy. That was until the release of Chat GBT in late 2022.

It blew the entire field wide open, turning automation from some niche trick used by some savvy companies into

06:20

the biggest thing since the internet. Generative AI models like chat GBT added to automation was like putting a V12 onto a bicycle for this automation space.

They made it possible to do more than just simple tasks. This was because these powerful AI models could handle much trickier stuff that used to require a human brain.

Instead of just updating

06:36

a spreadsheet row automatically, these automation platforms can now use the power of Chat GBT to do things that only people could have done before. So platforms like make.com, which you'll be learning more about later in this video, enable us to automate things like write a whole post for LinkedIn sounding just like you.

Pulling out names, places, and

06:52

phone numbers from giant documents in seconds. Reading and figuring out if an incoming email is someone asking for a refund or just wondering where their order is.

Shrinking huge piles of info into short and easy to read reports, spotting things in the picture like identifying a product in a photo, or even creating brand new images and

07:08

videos from just a few words. So what chat GPT and the explosion of other amazing generative AI tools gave us was basically human intelligence on demand.

These AI models are kind of like having a super smart friend who can do almost anything that you ask them as long as you tell them clearly what you want. And using automation platforms, we can

07:24

easily set up these super smart friends into our systems and use these kinds of models in thousands of different ways. All you have to do is pick the right AI tool for the job, give it a clear instruction, call a prompt, and watch the magic happen before our eyes.

And that is how the AI automation industry was born. So, with that little history

07:40

lesson out of the way, let's get back to our original question of what is an AI automation. Well, it turns out that this field is so new that there isn't even an official definition for what an AI automation is.

So, here's mine. Just keeping it nice and simple.

An AI automation is a system that uses AI to automatically do complex tasks that

07:57

would normally require a human. So, the big difference between these old school automations that we've just talked about and today's AI automation is the kinds of tasks that they can handle.

Thanks to these recent advances in AI technology, we've gone from just moving data around and putting stuff in spreadsheets to being able to solve problems that need

08:12

thinking and creativity and decision-making. It's kind of like upgrading from a basic toy robot that only moves forward to a high-tech robot that can solve puzzles and move around the world.

So, when you learn AI automation, you are basically learning how to build digital workers that can do very powerful things for you without

08:27

ever having to lift a finger. This is why so many people are racing to pick up the skill right now before it's too late to stay ahead.

It's like the ultimate cheat code because you can build them to do exactly what you need, tailored to any kind of job or workflow. For example, for students, it's kind of like having a helper that can organize your study notes automatically.

Or for

08:44

employees, your AI automations can be like a buddy that handles the boring paperwork so you can shine on bigger projects. And for entrepreneurs, it's like having a team member who runs your business tasks while you sleep.

And the great thing is that they all cost way less than hiring extra people. And they don't need breaks or vacation days, and they don't grumble about doing the

08:59

boring stuff. So, I'm sure you can see why businesses and students and anyone looking to save time and money are so excited about this technology and why anyone who knows how to build these systems is instantly 10x more valuable.

Just imagine automating something simple in your life or work like sorting emails

09:15

or scheduling tasks and how much time you'd be able to save to focus on what really matters in your life. Now, before we dive deeper, it's important to understand that AI automation is a super broad term these days, covering wide ranges of different systems and applications that can be

09:31

built with AI. This is largely due to the rapid advances in areas like AI agents and AI tools.

So, over the past 2 years, as I built my own AI agency, Morningside AI, and helped thousands through my communities to do the same, I've had to create a clear system for making sense of the chaos that is the AI

09:46

automation landscape. And I really want to share that with you today because it's been super helpful for me and many of my students to be able to figure this space out and at least have some mental buckets that you can put things in.

Here's the three different categories that you need to keep in mind. And please stick with me.

This will all make sense in a second. So firstly, we have conversational AI.

These are systems

10:01

that chat with people handing back and forth conversations. It's kind of like having a friendly robot that talks to customers for you.

These kinds of chat bots can be found on things like websites and answer questions or they can be voice agents that pick up phone calls. These used to need real people to talk, but now AI can automate these kinds of conversations.

The second

10:17

category is AI tools, and these are systems that use AI to do a specific job when a person asks them to, and it's mostly to help workers get more done. For example, I can make a custom AI tool that takes a link to a cool blog post that I found, grabs the info from the web page, does extra searches on the

10:33

topic, and then uses something like chatbt to write a new beta version for my own blog. And third and final is AI workflow automations.

These are systems that do a whole series of tasks by themselves, starting when something happens, like a trigger or on a set schedule like once a day. They use AI to

10:49

make decisions that used to need a human brain. It's kind of like having a smart robot manager that runs the whole process for you.

For example, an automation can call customers of an online store 14 days after they buy something using an AI voice agent to ask for feedback and review all without you having to do a thing. So, you set it up

11:04

on that trigger of 14 days after purchase, then execute this workflow. Now, those three categories might be a little bit confusing if you are completely new to the space, but don't worry.

In the building section of this video, we're going to be creating three automations which integrate each of these different types, and you get to see them in action, which will make it super clear. So, AI automation is

11:20

essentially an umbrella term under which all of the exciting stuff in the AI space is happening right now. However, when people talk about AI automations, they are typically referring to the last type that I mentioned.

So, AI workflow automation, and these are what we're going to be building later in the video. And an automation refers to one chain of steps that uses AI in various ways to do

11:38

certain tasks. And if I'm honest, this last category of AI workflow automation is really the most powerful because it can incorporate all elements of AI automation like agents, conversational AI, and tools in order to build end-to-end processes that are much more valuable than if they were just alone.

So, what I'm saying is that what you're about to learn is the key skill that

11:54

lets you do pretty much anything in the AI space these days. It is the foundation for building systems that save time, make money, or just make life easier, whether you're at school or work or running your own show.

And after years in the game, I can tell you that it's been one of the most valuable skills that I have ever picked up.

12:10

So, now that you understand what AI automations are, let's take a little peek under the hood and see how they actually work. So, don't worry if this sounds tricky.

I've been breaking down this kind of complex AI stuff for years now. So, I'm going to make this super easy to understand for you.

So, you can think of an AI automation like a facto's assembly line, right? There's different

12:25

stations and they're all working together to build something awesome from start to finish. It's kind of like having a a team of little robots, each with a special job passing the project along until it's done.

So, let's go through the five key parts to make this magic happen. Firstly, we have the trigger.

This is the very first step of an automation. You can think of it as

12:41

the facto's start button or the whistle that says, "Let's go." It's what kicks everything into gear. It could be something like a new email popping into your inbox, a form being filled out on a website, or even a specific time of day.

This is more of a a schedule. Secondly, we have a filter.

So, not everything

12:57

that starts in the automation should keep going through it. So, a filter essentially checks if what came in is the right stuff to work on.

It's like how a factory worker does some kind of quality control and checks that if the materials that they've received are good enough to use in the final product. If they are not then they get tossed out and if they are then they move forward

13:12

to the next part of the sequence. You can think of it kind of like a bounce at a club where only the important stuff and the good things that you want inside the club or in your automation are allowed through.

Thirdly, we have actions. So this is where the real work gets done.

Actions are the steps your automation takes like the different stations in a factory where each one

13:28

does a specific job. So for example, your automation might send an email, update a list or create some kind of report.

Often there are going to be several actions one after another just like a product moving down the line getting built bit by bit in order to achieve one of these outcomes. Next we have the intelligence layer.

So this is where the AI magic shines. This part is

13:44

like having a super smart robot on the assembly line that can think, analyze, and make decisions on the spot. And you can tell it how to think using prompting.

The AI inside your automation can look at each task, figure out what's needed and adapt based on the context you provided, like deciding how urgent something is, writing a custom message,

14:00

or pulling out key info from a big mess of data. These intelligent steps can go way beyond just following some kind of preset rules that we saw with old school automations.

Next, we have the format. So, just like items in production line may need some sanding or trimming before another piece can be added on top, the data in our automations often need some

14:16

kind of adjustments along the way. And this is where we use a formatter to prepare things for the next step.

And finally, we have the output. This is your finished product, just how Carac Factory packages up the final item and ships it out.

Your automation delivers the completed work at the end of the sequence. This could be a message sent to your team, an updated file in your

14:32

system, or a finished document that's ready to go. At the end of the line, it's where everything comes together.

It's like after making pizza and putting all the toppings on and prepping the dough, you're finally getting out of the oven. It's hot.

It's ready to eat and ready to go. So, let me show you how all of these parts work together with a real example that anyone in a job can relate to.

Say you're an employee who wants an

14:48

automation to handle incoming customer emails while at work. So the trigger is going to be when a new email lands in your company's support inbox.

The filter is going to check if it's something important like if the email mentions urgent or problem then it will pass it along to the next step. If not, it will end the sequence there.

Thirdly, we have

15:03

the intelligence layer which is going to use AI to read the email, figure out what it's about and then helps to draft a helpful response. And in this case, the intelligence layer also acts as a formatter which is essentially packaging the response and making sure it's in the format we want.

Then the actions in the automation can actually send that reply and alert the boss on Slack if needed.

15:20

The output logs everything neatly in your system and marks it as handled. So that's just a simple example, but these same building blocks are used to automate much bigger things too, which is why AI automations are such a game changer these days.

They combine these parts to create truly smart systems that can save time and boost efficiency,

15:36

whether you are you're juggling tasks at work or crushing it in school or running your own business. So, now that you've got a handle on what AI automations are and how they're built, let's take a quick look at the tools that make all of this possible.

Now, don't worry if this sounds a little bit too techy. Soon, these tools are

15:51

going to literally be second nature for you, and it's going to feel so easy to do. Creating automation starts with picking a main automation platform.

These platforms are typically called workflow builders, and they are the command center of your automation factory. They give you a blank canvas to design your automation on using the

16:06

building blocks we talked about. Some of which are going to be powered by AI like chatb.

Popular workflow builders include make.com which we're going to be using later in this video in the tutorial section. We have Zapia which is great for quick setups and we have NAM which is perfect if you want a bit more control.

They are the brain of your operation and they're really just

16:22

controlling how everything fits together. So workflow builders don't work alone.

They are essentially a place to hook up all sorts of other tools to get the job done. So these are the categories of tools which you can connect to your workflow builders.

Firstly, we have databases and spreadsheets. So for storing information and data, you can use things like Air Table or Google Sheets.

You can think of

16:38

them as basically the filing cabinets where you keep all your data neat and tidy so that you can save new things to the database or you can pull it into your automations as needed. Secondly, we have communication tools for sending messages.

Things like Slack or Google. They're like kind of walkie-talkies that can pass information around automatically for you.

Then we have AI

16:53

models. And these can add that smart human level thinking like open AI's chat GPT which you can think of kind of like a genius buddy who solves problems for you when you give him the instructions that he needs.

Then we have scheduling tools which can handle time and meetings things like calendarly or Google calendar and these are essentially like your personal planner to keep things

17:09

running on schedule. Then we have forms and intake too.

So this can collect information from people where you have things like type form or tally. These are essentially input forms and triggers for your automations when someone fills them out.

So as your automation goes through it's like calling up all your friends on a group project. Each one is bringing their own special knowledge or

17:24

capabilities to the table in order to help you to get to the end of that sequence and finish off the automation to create the final product. By building out a workflow, you are basically the boss at the factory.

Once you know what each tool can do, you can mix and match them to work together smoothly along this assembly line. It's kind of like building with Lego blocks.

You can just

17:40

like snap the right pieces together to make something awesome. You get to decide which tools you connect in what order and how AI can make the whole system smarter as well.

Which AI models do you use? Gemini, do you use OpenAI?

to use complexity for searching. There's so many different decisions you have to make.

Uh but it's really cool to be able

17:55

to flex your kind of creativity in order to solve these problems. It's really like a a new age form of problem solving, which is why I love it so much.

I say, I know I need to take this and get to this. How can I use AI to do that?

And really forces you to explore what's out there in terms of AI tools these days. So, what kinds of workflows can we build?

So for students, for

18:11

example, you could build a study material organizer that automatically summarizes lecture recordings for you, creates flashcards from notes, and schedules review sessions based on your exam dates. An employee could improve their workflow with a meeting assistant that records and transcribes their meetings, and then generates action item summaries off that and updates project

18:28

management tools. Entrepreneurs could design a lead qualification system, which automatically qualifies leads, calls them with an AI voice agent, and then sends them a custom proposal to automate that whole process.

And that is exactly what you'll learn how to build by following along with me in this next section. So, I'm going to show you how to build out a lead qualification

18:43

automation step by step, starting simple and getting more complex as we go. We're going to be focusing on this business-based workflow because these kinds of business systems offer the best opportunity for monetizing your new skills, which I'll go into depth on at the very end.

So, stick with me and you're going to be learning how you can start to make money immediately with

18:59

these kinds of skills. So, before we get into that, let's do a quick summary of this section.

So firstly, an AI automation is a system that uses AI to automatically perform complex tasks that used to require humans. Secondly, we build them inside of workflow builders that incorporate

19:15

integrations with tons and tons of other tools. And every automation has six key components.

Firstly, a trigger, what starts the workflow, a filter, conditions that need to be met, the intelligence layer, which processes info and makes decisions with AI, the actions, which are the tasks that actually get performed, formatterers,

19:32

which clean things up, and the output. the final result or deliverable.

So, if you're feeling unclear about anything we've covered so far, feel free to go back and rewatch some of those sections and be ready to join us in the next step when we begin building. So, it's really, really important that you do understand everything that we've gone over there because you're not going to have a solid

19:48

foundation to build your technical skills on top of, which we're doing in the next section. So, please, please, please, I've taken a lot of time to put all of this information very, very gradually together.

So, you must understand everything that I've gone over just here before we go into the next phase. If that's all good, then let's take a look at what we're going to build together.

20:05

Okay, so now that we have the foundational knowledge built that you need, we can now get into the second chapter of this video where we're going to be building three AI automations from scratch. We're going to be starting off with something super beginner friendly to get you started and then working our way up to a much more advanced and valuable one by the end.

Now, very important note is that each of these

20:21

automations build on each other. So, you have to be able to do the first one in order to be able to make the second and so on.

So, you you cannot skip ahead. I've planned this out very carefully to gradually layer on your skills.

So, it's all very intentional. So, please just trust the process.

So, over the next chapter, you'll learn almost all of the key skills you need to start building

20:37

your own AI automations from scratch and be able to tap into this enormous opportunity that is AI automation. The system that we're going to be building gradually over the next three sections is an AI lead qualification and proposal generation system for a business.

In the first tutorial, we'll be setting up the base using AI to automatically qualify

20:53

leads after they submit a form on their website. In the second tutorial, we'll improve that qualification ability by implementing an AI voice agent that can call the lead for more information.

And finally, in the third tutorial, we'll implement an automated proposal generator that can instantly create proposals for qualified leads using the

21:08

information collected on the phone call and in the lead form. The whole point of this is that as soon as someone's interested as contacting the business, they can kick things off with them immediately by getting a proposal in their hand.

So this process of qualification is a crucial part of running any business at scale because at the end of the day, not all people who come to a business are going to be a

21:24

good fit for their services. So for an example, someone may run an accountancy business, but they only choose to work with doctors.

They are specialized in helping doctors with their finances. But if a builder fills out their website form, then that lead would not be qualified, right?

Taking a call with them would be a waste of time because they are not a doctor. Therefore, to

21:40

stop wasting time, almost all businesses need some kind of qualification system. And for qualified leads, most businesses need to make some kind of proposal for them in order to kind of look at it and and see what they're proposing and then agree to that proposal.

Since most proposals don't lead to a deal, maybe 20% if you're lucky, this is a huge

21:55

waste of time and resources for the company. So, long story short, what we're about to build will solve a number of key problems for basically any business, making it extremely valuable for you to be learning how to do it.

And in fact, this is something you'll be able to go and sell directly to businesses when you're done. Here's what this process looks like without AI automation first.

So, a human sales rep

22:11

must constantly check for new form submissions from the website, review each lead's details, and evaluate if they're worth pursuing, research the company to understand how to help them, make phone calls, and deliver competent pitches, and then manually create custom proposals. That is hours of repetitive work per week.

And it could lead to potential leads slipping through the

22:27

cracks. If maybe they get the qualification wrong, they don't understand what the business is doing, they don't properly research it, or they're just too slow at getting back.

They may have to wait a whole day for the lead to hear anything back. So decreasing the time it's going to take for a qualified lead to hear back from them.

It's going to drastically increase their conversions. With AI automation, we can transform this tedious process

22:44

into something much more efficient, which is exactly what we're going to be building together. So here's how it works.

It's going to start as usual with a lead filling out the form on the website. Then the system is going to automatically qualify them using AI.

Then it's going to further research the company using AI also. Then it's going to send an automated phone call to pitch

23:00

your offer using an outbound AI voice agent. The system then saves the call outcome, summarizes the conversation, and generates a personalized proposal for them, all without a human having to lift a single finger and in a super scalable way.

This is the power of AI automation. We're taking tasks that used to demand hours of manual work and

23:17

turning them into workflows that run all by themselves. and we're going to build it all using several of the most popular tools from the automation ecosystem.

So, if you're as excited to learn this as I'm to teach you, then let's get started. And everything that you need to follow along with it, including all resources, templates, prompts, etc., is going to be available for free on my

23:32

school community. You can find it in the first link in the description.

You'll need to request to join. It will take 1 to 2 minutes to be accepted.

Once you're in, you can just search for the title of this video, and then you'll be able to find all the resources attached to it. So, that is how you can get all the resources to follow along.

Let's get stuck into it.

23:47

All righty, guys. Just to clarify things before we jump in, it's very important that you understand what we're doing in this first build, but also in the second and the third and how it all fits together and you get kind of an idea of where we're going.

Understanding this technology and then being thrown in a whole bunch of random things uh like Air Table and and Slack and make and all of these different terms can be a bit

24:03

confusing. So, I just want very quickly to give you a bit of an orientation before we jump into things of what we're building and why and how it works because I've touched on it before, but I don't think I did a good enough job.

So, I'm filming this after just to really make sure you guys are fully clear on what we're building here. So, and why it's valuable really as well.

So um as we touched on this is going to be an AI

24:20

qualification system um for inbound leads to a business. So easiest way to explain this is if we had a a website and say we had this book a free consultation form and we might be running ads to it.

We might just have that on a website and people just discover it by searching the web or or we're showing up in Google results. But

24:36

at some point, a business like the example I gave before, if you're an accountant and you only work with uh builders or plumbers or whatever the the example I gave, some people are going to fill out this form who are not necessarily the right person for you. Like the my accountancy firm that only works with plumbers, if a surgeon comes

24:51

in or if a a pilot comes in and fills this thing out and wants to get accounty services, we would ask them questions around, okay, well, what's your name? What what do you do?

Uh what sort of industry do you work in? And we'll collect this information here.

And as I mentioned before, not all of the people who fill out your forms, particular if you're running ads or or doing content

25:07

and it's driving traffic to this landing page as a business, not all of them are going to be people that you actually want to get on calls with because if you get on calls with them and they're not your ICP or the person that you could potentially work with, then it's going to be a complete waste of time. So, a qualification is a very essential part in in pretty much any business.

And what you're going to build is a very very

25:23

powerful system starting off with something basic like this. And then we're going to progress it to something a bit more advanced like this.

And then finally, something extremely valuable and powerful that you guys are all going to be able to build along and ultimately sell this to someone if you wanted to afterwards. So that's going to be covered in the monetization section.

But just to clarify sort of what we're going

25:39

to be doing here in this first build, uh you'll start off with building a form on a telly. So we're essentially replicating this kind of website form, but we're just going to build it in tally in this case.

You can embed these into different websites, etc. But just think of this as a form that people will get sent to.

So maybe they clicked on an ad and then they get sent to this. How

25:54

can we help? We're going to build out this form and then when people click submit on this form, it's going to send them into this basically a fancy spreadsheet.

And in this case, we're using Air Table. Air Table is like a really awesome platform which you're going to get to to use a lot within this build.

Uh but Air Table, you can think of it like a fancy spreadsheet and we're going to be able to map uh each of these

26:11

answers to a row in this database here. So every time someone fills this form out that we create, it's going to load it and add it as a row into this database.

And the cool thing there is that air tableable allows us to within make.com our workflow builder it allows us to set up events that when things happen in air table like okay let's

26:27

watch for new records that arrive and that's what this here is doing. So this trigger of our automation is going to be watching for new records that arrive in this and then it's going to take anything that arrives in here say this first row and then it's going to put it through this automation here.

And so this is going to be a very basic one

26:43

where we've taken actually the AI component out of this. Before Air Table really got advanced with the AI features.

What you're doing here is take the information from the Air Table. You'd use some chat GPT step here and then you'd be able to do things afterwards.

But in this case, I'm making it extra simple for you guys and that we're actually just using Air Table's built-in AI features. So down here, you

27:00

can see in the tutorial that you're about to go into, we will use Air Table to create an AI column here that's going to take in information like uh like here where they're talking about their their company, their budget, and about their needs. We can write a prompt in here and it's going to automatically fill out this row.

So as soon as a new person

27:15

arrives in this spreadsheet, this field that we're creating here is going to automatically qualify them and use AI and this prompt that we write here to analyze the information that they gave us and then output if they are qualified or if they're not. So we're doing the qualification with AI built into Air Table which is going to save us a lot of

27:30

time and messing around with jumping back and forth. So So that is basically the start of build one.

We will do immediate qualification with AI within Air Table and then we're going to if they are qualified we're going to send an email to them and because we're already doing the qualification here in Air Table it makes it a lot easier for us over at make.com because for this

27:46

particular node as you'll see in a second we're going to set it up so that it's only actually going to look for uh the rows or records in this table that have been AI qualified as qualified. So if this prompt here that we've written analyzes all the information in the row and then outputs qualified then and only

28:02

then will it pass it into here. And this is when we're going to be sending an email to the prospect and saying, "Hey, we're interested in in hopping on a call.

Here's a link to book in a call with us." But at the same time, it'll also send a message to our sales team in Slack and say, "Hey, look, we've just had a new qualified lead." So, this is

28:17

very much a a minimized version of what you could do uh maybe a year ago. But because of the Air Tables AI features, uh we can do a lot of the heavy lifting here and writing a prompt that's going to analyze what the person has filled out here, aka are you a are you a builder or are you a pilot?

And if they

28:32

were a builder, in the case of the example I just gave, it's we're going to be using a different example in the build. Um, but it would come out as saying qualified here.

And because it says qualified, then it would automatically trigger uh the rest of this automation, which is sending an email to them and taking the next step and saying, "Yes, we are interested in talking to you further, making sure that

28:48

we're not talking to anyone that we shouldn't be talking to and at the same time letting the sales team know." So, that is build one. Um, that is how it all fits together.

And I'm going to be doing these little updates just before each build so that you guys are 100% clear on what we're going to be doing.

29:03

To get started, we need a way to gather our leads information. While there are several great options for this build, we'll be using Tally, an easy to use form builder.

Just sign up for free or log into your account. Then we'll click on new form and give

29:18

our form a title. Anyone using the form will see this title.

So, we want it to make sense for the user. So, in our case, we can name it, how can we help?

I want to build it from scratch. So, I'll hit enter.

The way we create forms in tally is with building blocks. So, if we click the plus button, we can select the type of block we need from this list of

29:35

input options. We'll start with a short answer input for the user to write in their first name.

And we'll do the same for their last name. Notice how we're adding a label above each input, which tells the user what to write in each of these.

Then, we'll select the email block for the user's email. Add in a

29:51

phone number block. Then, another short answer input for the name of their company.

Since their budget is a number, we'll add a number block to grab that. And finally, we'll add a long answer input where leads can describe their specific needs.

Now that our form inputs are ready, we can

30:07

customize the form's appearance to match our branding. We can adjust elements like the background color, text color, button color, and the accent color.

Feel free to choose whichever colors you'd like. Now, we can go ahead and publish the form.

30:23

When it's ready, we'll see a sharable link. Let's copy this and actually visit the published form.

It's looking great and ready to use. In the real world, you could either use this as a landing page or embed it into a website.

Whenever a new lead fills out this form, we need to be storing that response somewhere. For

30:40

our build, we'll store our form response inside Air Table. You can think of Air Table kind of like Google Sheets on steroids.

It's your database that not only stores information, but can process information and if you push it far enough, can even be used to build out more complex apps. So sign up for free

30:56

or sign in if you already have an account. On the home screen, you can see that we could use AI to build things or start with a template like one for a marketing campaign or a project tracker.

But for our build, we'll start by creating a base from scratch. Base is just Air Table's name for a database.

31:12

Again, it's really similar to a Google sheet or Excel sheet, but can be layered with complexity. We'll name it lead base.

We'll reference this name later when setting up the connections between our tally form and air table and later between make and air table. For every field in our form, we need a matching

31:27

field in air table. So, we'll add fields for our leads.

First name, last name, email,

31:44

phone, company, budget will reference this to determine a lead's qualification. notes, details they submit about their needs created on when the lead was added to this base.

32:01

Importantly, we need to know if our lead is qualified or not. Ultimately, lead qualification can be more nuanced, but for our needs, we'll qualify leads whose budget is 10,000 plus.

By using Air Table's AI assistant, we can prompt it to create a custom field for us by asking it to create an AI field called

32:17

qualification that sets the lead as qualified or not qualified based on whether the budget field is greater than or equal to $10,000. Now, Air Table will use the power of an LLM open AI at the time of this recording to help with creating this custom field.

After

32:33

processing, the AI fields modal will appear. Since we want leads to be autoqualified, we'll enable the automatic generation option.

If you encounter errors about missing or invalid fields while saving your AI prompt, it's typically because the system can't clearly identify which

32:49

fields you're referencing. Double check that all fields are valid and properly linked to the correct data in your base.

We want our database not just to autoqualify leads intelligently, but also to generate a descriptive message about each lead that we can share with

33:04

our sales team. So, we'll use the AI assistant again to create a field that generates this message based on the lead's information.

After double-checking that this new prompt looks good, we can save it and move on to the final field. We'll add a date field called contacted on that will

33:21

store the date when we contacted the lead through our automated email. In the make workflow, we'll add the date to the field from make at the very end of our workflow.

Finally, we'll clean up this base by renaming the sheet to lead contacts and set up each record or row to be called a lead. Now that we have

33:37

all the necessary Air Table fields that match our form fields, we can connect Tally to Air Table via Tally's Air Table integration. Just name the connection, select the database and table you're syncing to, and map each of the tally fields to

33:54

their respective data fields in Air Table. After saving this connection, we'll see a confirmation that the integration was successful.

Now, let's test the connection by filling out the form with some dummy data,

34:09

making sure to set the budget to $10,000 or higher. When we submit, we should see the new lead record populate in our Air Table database.

And voila, it's showing up perfectly. Since the budget is over $10,000, it's automatically marked as

34:26

qualified. And we've got an AI generated message ready to send to our sales team via make.

With our lead form synced up to our database, we're ready to start building out the automation workflow. This is what we're about to build out.

We'll have a module that watches our Air Table database for new leads. And when it finds one, it sends them an email via

34:43

the Gmail module, then updates that same leads air table row with the time we contacted them. It also formats and sends a message into Slack channel.

If you're not familiar with Slack, it's a professional group chat with channels dedicated to specific topics like marketing. With this plan in mind, let's head over to make.com.

35:00

Remember, this is our workflow builder, the assembly line where we'll be constructing our automation. You can get started for free or login if you already have an account.

Once inside, you'll see your dashboard, which displays info about your workflows, which are called scenarios in Make. You'll see how busy they've been and how much data they've

35:16

used. This helps you monitor your make plan usage, which is particularly important if you're on a free plan with usage limits.

You can also see if any of your scenarios require attention due to things like bugs and errors. The scenario tab is where you'll spend most of your time, and you can organize your scenarios into folders like you see

35:32

here. Before we start creating our own, I want to bring your attention to the fact that you have a bunch of templates to get started from and adapt to your own use cases.

And over time, you'll be managing the connections here in this tab, which shows you which external apps like OpenAI and Google you have

35:48

connected to from within make. With that brief tour out of the way, let's get into it and create a new scenario.

We'll start by setting up the trigger. After selecting the Air Table module, we'll configure it to watch records.

Since make doesn't yet know which

36:03

records to watch, we'll need to establish a connection between Make and our Air Table account. We'll do so with a token because as you see in this warning keys are deprecated.

So we just need to create a token and paste it here. To get that token go to your Air Table account settings,

36:21

navigate to the developer hub and create a new token named something descriptive like make token. When configuring the scope, set it to allow both reading and writing records with full database access for read and

36:37

write operations. I'm granting access to all current and future bases, though you may want to restrict this depending on your specific needs.

Now, click create token and immediately copy and store it somewhere secure. You'll never be able to see the

36:52

complete token again. Once we paste the token back into make, our connection to air table should be successfully synced up.

Now we just select the bas and table we want make to be watching and set the trigger field to created on.

37:08

The trigger field tells make which field to monitor for changes in order to determine when a record is new. By setting it to created on make will only pull records whose timestamp is later than the last time it checked for new records.

That way it only acts on newly added leads and doesn't rerun on old

37:24

ones. As for the label field, we'll set it to the company.

This field is just for display purposes inside makes scenario logs. It helps you identify the records that are being pulled in, but it doesn't change how the automation works.

The formula field lets you write an air

37:39

table style formula that filters records before they're passed into your automation. Make only processes records when this formula evaluates to true.

So, if a lead's qualification field is equal to qualified, they'll be pushed through to the next step of the automation. If they're not qualified, they don't pass

37:55

this step. If you ever need help writing these filters, you can ask an LLM like chat GPT or even have a chat with Make's built-in AI.

Once we hit save, we'll choose to start from now on. That way, make only looks at new leads created after this point and skips anything already sitting in Air Table.

If we hit

38:11

the run button, it should successfully find the newly created qualified lead and display that records details. Once you run this scenario, any qualified leads that were found will not be found again on rerun due to this created onfield.

You'll have to add a new lead each time you want to test this scenario. Again, the automation is

38:28

filtering for only new leads. Since we want to send qualified leads an email to schedule a call and notify our sales team about this lead via Slack, we'll need to set up two actions.

This is where a router comes in. It allows our scenario to branch into these two separate tasks simultaneously.

Without a

38:45

router, setting up tasks sequentially means if the email step fails, the Slack step won't run at all. We want to avoid these waterfall effects where one failure blocks other actions.

Using a router prevents this issue while making our scenario easier to troubleshoot and expand over time. On the first route,

39:01

we'll add the Gmail module and select the send an email action, which requires us to set up a connection to our Gmail. If you're using a non-personal email like mike@edge.ai, this setup is very simple.

You just log into your Gmail account here. But if you need to use a personal Gmail, one that

39:18

actually ends in gmail.com, the process is more involved. So, I'm going to ask you to be patient because there are quite a few steps involved to connect a personal Gmail account here.

If you're just watching and not building along yet, feel free to jump past this section. Same goes for those of you who are using a non-aggmail.com email.

But

39:35

if you do need to connect your personal Gmail account, here's what that looks like. As you can see, if we pop open these advanced settings, this requires two key pieces of information.

a client ID and a client secret. These act as the key that allows make to unlock access to your Gmail account.

Make provides

39:51

instructions for how to generate these, which you can find by clicking on this guide link. At the time of this recording, it links to the help center.

From here, click into apps documentation, then click on communication and scroll down to Gmail. If you're wondering why make didn't just link us to this page, so am I.

From

40:07

here, we'll click on create a custom oorthth client. And now we need to log into the Google Cloud Platform.

Once you log into Google Cloud with the Google account you want to send emails from, you'll click select project at the top left. From there, click new project, then give

40:24

the project a name like make and hit create. Once it's created, a notification will appear in the top right corner.

Click select project to open your new project. Now that we're inside, let's enable the Gmail API.

Go to APIs and services

40:41

and click enable APIs and services. In the search bar, type Gmail and select the Gmail API from the results.

Then click enable. Next, we need to tell Google who is requesting access and what kind of data

40:57

is being requested. So, we'll click on Oorth consent screen and hit get started.

To fill out the app information, we need to give the app its name. So to keep things consistent, let's call it make like we did for the project name earlier.

We also need to tell it which email we want to give

41:12

these new abilities to. Now we select external for the audience type and next.

Under contact information, enter an email for Google to notify you about this project. Hit next.

Then agree to the terms, continue, and hit create. We now need to give make permission to

41:29

interact with Gmail on your behalf. So head into the data access tab and click add or remove scopes.

We're adding Gmail API scopes. So, search for Gmail and then select the following scopes.

Both scopes to read, compose, and send emails. Manage drafts and send view your

41:46

emails and settings. View your metadata.

Add emails to your mailbox. Send email on your behalf.

See, and edit labels. With all those labels selected, hit update.

Then, scroll down and make sure to save these scopes. Next, let's configure the branding settings.

Scroll

42:01

down to authorize domains and add both make.com and integromat.com which is the old name make used to be called. Then save.

After that head to the audience section and click add users where you'll add your Gmail address and hit save.

42:18

Now it's finally time to create the actual credentials that make will use. Go to the client section and click create client.

For the application type, choose web application and name it make. Scroll down to authorize redirect URIs.

Click add URIs and paste the exact

42:35

redirect URI provided by Make. You can find this in their Gmail integration tutorial page that I showed you earlier.

Click create and you'll be given your special client ID and client secret.

42:51

You will copy both of these and paste them into the field in makes Gmail module. Now we can click sign in with Google.

This sign-in window will appear. Select your Gmail account.

Grant the requested permissions and hit continue. Once

43:07

that's done, your Gmail account should be successfully connected to make. I know that was a lot and I wish it was simpler too.

But the good news is we can now start sending emails from that Gmail account. So, let's now create the email template.

We'll set up a subject line using the lead's name from their Air

43:23

Table record and whatever engaging text you want to add in here. For the content of the email itself, we can address them by name and say, "We received the request about their company and would love to discuss their goals on a call they can book via our scheduling link.

For this, we will be using a handy tool

43:40

called Kalanley. If you're not familiar, Calendarly is a scheduling tool that lets people book meetings with you.

You connect it to your calendar, set your availability, add conferencing tools like Zoom or Google Meet, then create event types like an introductory call and share a link for people to book these meetings with you. So just set up

43:56

a new account if you don't have one or log in and we'll create a new event type and call it intro call. This event will have a 30-inut duration and the location will be Google Meet.

Keep in mind you may have to connect a conferencing tool if you're using something like Zoom. Now

44:11

we can set our availability for this intro call event. I already have mine set up, but to do that, you just head to the availability tab and go to calendar settings.

Here you can see I already have a couple calendars synced up, but to add a new one, click connect to calendar account and select your

44:26

provider, such as Google. Then just choose the account you want to sync and log in.

Again, I've already set this up, so I'll close out of this and head back to the event. For each event, we can set the availability for that event type.

I may want to change the time I'm free on Friday, either a specific day or every

44:43

Friday. When I apply these changes, that will update the availability for just this event type and does not apply to other events I may have set up in Calendarly.

Down in more options, I can do things like add an event description where I can tell the person booking that I'm excited for our chat. Once we hit

44:58

save changes, the event is ready to share. So, let's copy this link and head back into make and paste it into the body of the email we're sending to our lead.

Finally, we of course need to add the recipient of this email. So, we'll fill that in with our lead's email address.

With that set up, we can move on to our other route below. Ultimately,

45:15

we're going to be posting a message to a Slack channel about our new lead. But if you don't use Slack and don't want to go through the steps of connecting another external tool, you could sub this out for another Gmail module where we send an email to our sales team letting them know that we have a new qualified lead and give them the lead's name and other

45:31

relevant info. Then set the recipient email to whichever test email here.

But since I want to show you how to use a bunch of popular tools, let's look at setting this up to post to Slack. We'll choose the Slack app and select the create a

45:47

message action. If you already have a Slack organization, you can simply connect make to that module.

Now, if you don't, let's quickly walk through creating a new one. You'll head to and sign in with a new account, then click create a workspace.

Give it a name.

46:02

Here, I'm just adding the name of our mock agency that we're qualifying leads for. Then just run through these steps to confirm your name.

Skip adding new members for now and start with the free version. Now we're in our new Slack organization.

As you can see, there are channels here for different topics and direct messages down here. We'll add a

46:19

new channel called marketing because this is where we'll be posting our Slack message into for the entire sales team of our imaginary agency to see. With these ready to go, we can head back to make and connect the Slack module to our Slack or by allowing access.

46:35

We'll tell the module where to post the message by selecting from a list of our public channels, specifying the marketing channel we just created. Now, what should it post to this channel?

It should post the AI message from our Air Table lead. So, we'll add that in here.

If we run the scenario, it will work, but the message will look messy because

46:51

it's technically a collection, which just means it includes extra stuff that we don't want or need to display here. We can solve this by inserting a formatter module just before we post to Slack.

We'll select a text passer, which will do a pattern match using a regular expression to pull just the clean

47:06

message text out of it. While this might look complicated, think of it like a smart highlighter that scans the text and grabs only the part we care about, the messages value itself and not the messy metadata.

If you ever need to use a regular expression or rejects in the future, just ask an LLM to draft one for your use case. Now, we tell our text

47:24

passer which text to pass, the AI message. With that formatter taken care of, we can head back into the Slack app and tell it to create its message based off that freshly passed text instead of the original messy version.

As a final step, we want to circle back

47:40

to Air Table and update our lead record with information about when we auto contacted them. After the Gmail module, we'll add an Air Table module and use the update a record action.

We'll configure it to return to the same lead base and contacts table targeting the ID of the record that just pushed through

47:55

this scenario. We'll set the contacted onfield to the current date and time by using the now expression which simply tells make to insert the exact moment the automation runs.

If we run the scenario and head back into air table, we'll see the date was successfully added to the contacted

48:11

on field giving our sales team context about when the lead was auto that email. Our scenario is all set up and working great, but up until now, we've only been running it manually by clicking the run once button by toggling on the schedule.

We can configure our scenario to run on autopilot at regular intervals such as

48:29

every 15 minutes or on a specific day of the week or month. You can even set a custom schedule using advanced scheduling using time ranges with start and end dates if you only want it to run during a certain window.

Once you're confident it's working, just flip that schedule toggle on and your automation will run in the background while you

48:45

focus on more important things. Of course, you'll want to frequently save the scenario.

And if you ever need to revert to a previous version, you can revert to that version from here. Scenario inputs are useful for more advanced use cases, like when one scenario's output becomes another

49:00

scenario's input, but that's beyond the scope of this basic build. You've also got your scenario settings, a place for notes, an auto aligner if your workspace gets a little messy, and even a little animation that explains the workflow.

49:16

Finally, there's a quick reference to every app and module used in your scenario, helpful for getting a bird's eye view of your automation. And so, that's build one explained, just a very basic AI qualification based out of Air Table.

But to give you an idea of where we're going with this, in the next build, build two, we're going to be adding in a voice agent here. You may

49:32

have heard about voice agents before, but they're are a really a really exciting uh area of the AI space right now. You have two main types.

You have inbound and outbound. And uh inbound is when you can set up a phone number and people can call.

When people call that number, then they get to talk to the AI directly. Outbound voice agents is what

49:48

we're going to be building here where we can initiate a call using our automation here uh to send a call out to someone and say my phone starts ringing and I can pick it up and I'm talking to the AI that we've created here. So this is really really cool stuff and super powerful.

But the reason we're doing this is because the as you will have

50:04

seen on the form that we set up, we have quite a limited amount of information that we're collecting from the business and this is for good reason. You don't want to put too much information here uh or it will decrease the number of people who fill it out.

So sort of having a lean form and then if we initially qualify them here in the second build we're going to expand it so that we can

50:20

actually once they are qualified as we've explained here the AI and air table is going to determine preliminarily if they are qualified then we will use our voice agent to actually call the person and ask for more information and we'll walk through a script basically and ask them hey tell me a bit about your business and your needs what are you hoping to get out of

50:36

us da da da da basically collecting a bunch more information that can then be used for even greater and more accurate qualification. So here we have our Vappy voice agent which is going to call them and then we're going to get the data of that call, the transcription of that call and Vap is actually going to be able to analyze it for us to determine

50:52

if it was a successful qualification or not. And then we have to build the automation to handle a few different cases because of course not everyone's going to pick up the phone.

So here we create a route that handles if they answered the phone and then based off the information were they interested or were they not interested and if they didn't answer we have some other things that we can do to handle it here and

51:08

update the air table and things like that. So, it's essentially build two is adding on top of what we've already built on build one uh where we have just a basic qualification and sort of send them an email to book in a call or to let the sales team know.

Here, we're trying to do even deeper qualification and making sure that our sales team

51:24

really isn't getting on any calls they shouldn't be by sending out a voice agent to collect more information for us to do an AI pre-qualification for this business's lead. So, let's take this automation to the next level by adding AI voice calls to qualify leads even more effectively.

We've already built a strong foundation, a system that

51:39

identifies qualified leads, follows up via email, and notifies your team. This alone puts you ahead of the curve.

But what if we could make the system even more responsive, more conversational, more human even? That's where an AI voice agent comes in.

By weaving it into our workflow, we can automatically call

51:55

leads, gauge their interest through real conversation, and tailor our follow-up based on their responses, all without lifting a finger. This is where automation becomes more than a task runner.

It becomes a teammate. It's the closest thing we have to a human conversation at scale.

And when it comes to leads, timing and tone matter.

52:10

Research shows that responding to leads immediately increases success rates by 7 to 9x. In this next section, we'll integrate VP, a popular AI voice agent service, into our workflow.

First, we'll set up our voice agent. Then, we'll configure make to have it call our leads.

Finally, we'll enhance its effectiveness by feeding it custom

52:26

research about each lead, enabling more targeted pitches. So, what exactly is an AI powered voice agent?

You can think of them like a version of Siri or Alexa, but one that's specifically designed to have natural phone conversations with people. Depending on your needs, it can either make calls for you or answer calls.

These agents can do amazing

52:42

things like follow up with potential customers, schedule appointments, answer common questions, collect feedback, or conduct surveys. There are several platforms that offer AI voice agents like 11 Labs, great for highly realistic voices, and Vappy, known for being fast and affordable for our automation.

We'll

52:57

be using Vapy because it's beginner friendly, cost-ffective, and integrates well with other tools. If you haven't already created an account, go ahead and do that first.

Once inside, we'll go to the assistance tab and click create assistant. This starts the process of configuring our

53:12

assistant. You can start from a template like a lead qualification specialist, but we'll create our own from scratch and name it Ben.

When choosing which provider and model to use, you'll want to balance ability with speed and cost. So, keep the price in mind and the time it takes to

53:28

actually respond, which we call latency, which you can see next to the model options. For now, we'll go with the cheapest and fastest option, which is GPT4.1 Mini at the time of this recording.

The most crucial element is the prompt, which defines your agents behavior and objectives during the call.

53:43

First, we specify the opening greeting message that the agent will use when calling someone. We'll open up the call asking if it's a good time to talk about the lead's business needs.

The actual prompt is much more detailed and I'll paste in one I wrote earlier. You can find this prompt and everything

53:59

else you'll need to follow along in the first link in the video description. Let's take a look at what it includes.

As you can see, we start off by defining its identity and role. A voice assistant representing Edge AI and AI automation agency.

54:16

then specify its goal to pitch our services and determine whether the lead would like to receive a proposal. We then set its tone and behavior.

A knowledgeable rep who is curious but not pushy and who uses casual natural language with words like uh and mhm.

54:32

This section helps humanize your assistant. Finally, we outline the structure of the call that it should replicate.

The conversation flow starts with introducing our AI services, then moves to understanding the prospect's needs through targeted questions. After acknowledging their responses, we pivot

54:48

to offering a proposal. If pricing comes up, we defer to the proposal for details.

The call ends either by confirming they'll send the proposal over email or with a plight goodbye if they decline. It's best to outline the steps as clear bullet points like this, but keep them focused.

Too many steps

55:04

can overwhelm the AI assistant. In our prompt, we included context about the company the assistant is representing.

For our needs, this is sufficient. But if our call was more involved, like a customer support agent who is receiving hundreds of nuanced calls, we could add files into its knowledge base that it can reference, like all of our policies

55:20

and procedures. But keep in mind, this will increase the response latency since the assistant will need to make round trips to these files as it generates its responses.

So, if you can get away without adding supplemental files, your agent will be more performant. Our assistant is almost ready to start making calls on our behalf.

But before

55:35

we publish it, let's add some extra configuration. As we scroll down, we'll see that we can tweak the transcriber settings, which we'll leave alone for now.

We can also select different voices depending on the personality we want our agent to have. For our agent, we'll go with Vap's Elliot voice, but feel free to try out different options to feel out

55:51

which one is best for your use case. You can also add background noise to the call.

Adding ambient sound makes conversations feel more natural and helps mask any brief delays that occur while the agent generates responses instead of awkward silence. Callers will hear realistic background sounds.

Next,

56:07

you'll see the tools section, which unlocks powerful ways for your agent to take action during a call, like sending data to your CRM, triggering external workflows, or running custom logic. This is where Vappy really shines for more advanced automations.

However, since it's beyond the scope of this tutorial,

56:22

we'll leave it untouched for now. What we will be using is the call summary.

Vari automatically transcribes the call and the prompt inside here tells it how to summarize the call based on that transcript. If we open it up, we'll see there's already a built-in prompt that tells the assistant how to create its call summary.

While this default prompt

56:39

works decently, we can make it even more valuable for our make scenario by customizing it to capture additional details. Here, I'm pasting in a new prompt telling it to summarize the call in a couple sentences with details that will help out with creating a proposal later on.

Since we'll add this call summary to our lead's air table record,

56:55

capturing detailed insights here is valuable. We also want to tell Vappy how to judge if a call was successful or not.

There's already a decent default prompt here, but we'll customize it for our needs by telling it to judge success solely on whether the lead wants to receive a proposal. We'll evaluate success with a pass fail, which simply

57:11

gives us a true or false value that we can use in our automation workflow. While there are additional advanced options here, the final step we're concerned about is the voicemail detection.

Let's make sure we toggle that on with VP as the provider. This is helpful for our make scenario since we

57:26

need to accurately track whether the call was actually answered or if it just went to voicemail. We're now ready to hit publish and test our newly created voice assistant.

Hey, it's Ben from Edji. Is now a good time to talk about your business needs?

Yes, it is. Great.

57:42

I'm following up on your submission because we build custom AI agents and automations for companies like yours. anything from customerf facing bots to internal workflow optimization.

I'd love to understand a bit more about what led you uh to reach out. What sort of processes or tasks are you looking to

57:59

automate? I'm looking to implement custom AI agents to facilitate learning on an e-learning platform.

That sounds really interesting. Um so you're looking um to enhance uh the learning experience on your e-learning platform with AI agents.

Are there any specific

58:14

challenges or bottlenecks you're facing right now that you hope these agents could help with? Uh, yeah, just streamlining the learning of the content library on my platform.

Got it. So, you're looking to streamline how users access and learn from your content library.

That makes a lot of

58:30

sense. Thanks for sharing that.

Would it be helpful if I sent a quick proposal your way showing how we could support your goals? Yeah, that'd be great.

Perfect. I'll send that to the email we have on file.

Thanks again for your time and I look forward to helping you with

58:46

your project. Great.

Thanks. Bye.

You're welcome. Have a great day.

Bye. Great.

Our assistant is working as intended. It's now the voice of our lead outreach standing by, ready to make calls on our behalf.

Before integrating this assistant inside of Make, let's walk through the new path a lead will take

59:02

through our extended scenario. We left off here, but we're going to extend this for the trigger.

Nothing changes here. A new Air Table record marked qualified still kicks things off.

Then our voice agent will call the lead using the number from their record. We'll add a short pause to wait for the conversation to finish before we analyze what

59:18

happened. We grab the call record from VAP.

Then we check if the call was answered. Answered.

Great. We log the call summary and mark the lead as interested or not in Air Table.

No answer. No problem.

We fall back to emailing them a link to schedule a call

59:33

and alert our sales team. Since we're going to be logging call summaries and interest level about our leads, we'll just need to head to our lead base in Air Table and add those.

We'll add a long text field called summary, a checkbox for interested, and while we're at it, let's add a new date for

59:50

when a proposal was sent on since in the final phase of this buildout, we'll be generating and sending proposals to these interested leads. Back over in make, we're ready to extend the lead qualifier scenario that we built earlier.

To keep things clean, let's just clone what we built and extend from there. Name it lead qualifier plus voice

00:07

agent. And now we're ready to continue building.

The beginning step remain the same where we watch for new leads. Then we'll add the VP module to create an outbound phone call.

We will need to set up the connection to our VP account,

00:24

which means we'll head back into that dashboard, click on API keys, then copy your private key. If you don't see one in there, just go ahead and add a new one.

Give it a relevant name. Decide if you want to restrict it to only work on certain sites.

Or if you only want it to work with specific assistant, then

00:39

create the private token. You'll see it pop up here.

Then just copy it. But make sure to save this somewhere safe since you won't be able to view it again.

Pasting the key into the VP module should set up our connection. Then we can configure our call.

We'll just fill in these fields, giving it the assistance ID, which you copy from the

00:55

top of the Vappy Assistant page. We provide the assistant with the lead's phone number pulled from their air table record.

But what number is our assistant calling from? If we head back into Vappy and click on the phone numbers tab, we can create a new free phone number with Vappy, specifying the area code to call

01:11

from. Note that at the time of this recording, only US area codes are supported by Vappy.

Since we're building this assistant for learning purposes, we can simply use a US area code like 223. However, if you plan to deploy an assistant for production use and you or

01:27

your client are located outside the US, you'll need to import a number from something like Twilio or Vonnage. Okay.

Once that number is created, we can copy its ID, not the number itself, and paste it into the makevarpy module. And with that, our voice agent is set up and ready to make calls triggered by our

01:42

make scenario. Once the call is sent, we need to wait a bit before we check what happened.

Calls take time to connect, ring, and potentially have a conversation. If we try to fetch the result too quickly, we could just get an empty response or an error.

So, let's add a short pause. Insert a tools sleep module.

Set the delay to 300 seconds.

01:59

This little breather, 5 minutes exactly, gives Vappy time to do its thing before we peek in to see how it went. To ensure things are working so far, let's add a new lead to our Air Table base.

We could add them via the form, but it's quicker just to duplicate an existing lead in order to create a new one. Now, we'll head back and make and run the scenario

02:15

to test that it's all working. Hey, it's Ben from Edgeai.

Is now a good time to talk about your business? Yeah, it is.

Great. I'm following up on your submission about exploring AI solutions.

02:32

We build custom AI agents and automations for companies like yours. Anything from customerf facing bots to internal workflow optimization.

I'd love to understand a bit more about what led you to reach out. What sort of processes or tasks are you looking to automate?

Yeah. So, I'm just interested in um

02:49

seeing how we can build out some maybe AI powered workflows to help the operations in my company. That sounds interesting.

AI powered workflows can really streamline operations. Are there any specific

03:04

bottlenecks or manual tasks you're hoping to reduce? Hoping to help my uh learning management system run more smoothly.

Got it. Optimizing your learning management system can definitely enhance efficiency and user experience.

Uh,

03:22

thanks for sharing that. Um, would it be helpful if I sent a quick proposal your way showing how we could support your goals?

Yeah, definitely. Perfect.

I'll send that to the email we have on file. Thanks again and I look forward to your feedback.

Thank you.

03:39

You're welcome. Have a great day.

Now comes the part where the automation becomes observant. We're not just automating calls.

We're building a system that pays attention to what happened. Did someone pick up?

Did the assistant get the job done? Was the lead interested?

To answer those questions,

03:54

we need to ask VBY for the results of a specific call. And to do that, we'll use an HTTP module to make a request for that call record.

To better understand this module's function, let's explore what an HTTP request is and how it fits into the broader workings of the internet and how APIs fit into all of

04:11

this. I know it sounds complicated, but let's break it down with a simple analogy.

Remember earlier we learned how an automation workflow is like a facto's assembly line where each module is like a machine performing a step in the process? Well, factories don't operate in isolation.

Sometimes your factory

04:26

needs supplies, information, or services from outside its own walls. A real factory might call a supplier to order more raw materials.

Ask a logistics company where a shipment is. Verify parts.

Meet quality standards with lab testing. request a maintenance crew to check the temperature of a remote

04:42

machine. Simply put, a factory coordinates with external partners to handle tasks outside its own expertise.

Just like real world suppliers don't take factory orders shouted over a fence, external services need a structured way to receive requests. That's what an API or application

04:58

programming interface is. In our factory analogy, the API is like the official order form that your external partners use to process requests, which you would fill out to order some special machine parts to use within your factory.

Online services like Vappy, Air Table, or Slack all offer APIs that follow the same

05:14

principle. They give outside entities like make.com a structured way to send and receive information.

APIs give us a clear, consistent way to ask for something and get a predictable, reliable response. There are several types of HTTP requests that serve different purposes.

So, you could be

05:29

saying, "Get me this thing. Post or add this new thing.

Put this info where it belongs. Replace the whole thing.

Patch just this one part of the info, but don't replace the whole thing. Delete this thing." In our case, since we're about to be making an HTTP get request

05:45

inside of Make, we're essentially saying, "Get me this call summary. Here's what I want, and here's who I am.

I've got the proper permission to access what I'm requesting." While platforms like make.com and n10 provide a visual interface for building workflows, under the hood, they're actually making API

06:00

calls to connect with these external tools. So even though you're working with visual blocks, these modules use the same underlying language that developers use to connect services across the web.

Yes, this is no code development, but that doesn't mean code isn't running. It's just happening behind the scenes.

That was quite the detour, but an important one because it

06:17

gives you a firm grasp on how things work on a deeper level. Now that you understand what an HTTP request is and how it fits into the big picture, let's set this module up to go get the record for the call that was just made.

We'll set the method to get. But where are we

06:32

getting something from? Well, we're going to make a request to VP's API to get the call that just happened.

If we check out the VP documentation, we can see that we need to send our request to this URL where the last part of the URL is the ID of the call we're fetching. So let's copy that URL and paste it into

06:48

the HTTP module. Adding on the call ID from the vari modules output in the header section.

Here's where we add instructions like putting a label on a package by putting authorization in the name field. This just means we're saying we have the authorization to get this info and the proof is the value of

07:05

the header itself. Another way to think of that proof is that we have a key to unlock this special box that contains the information we're requesting.

That key is the API key that we generate from VPY. So in the value field, we'll type bearer.

Then paste that key next to

07:21

it. This roughly translates to the person who possesses or bears.

This key has permission to access what they are requesting. Finally, we'll say yes to passing the response.

This means the module will break apart the response into structured fields that we can

07:36

easily use later in the scenario. Let's make sure this is working by running this module only.

Here at the bottom, it's asking for the call ID that we want to get, which we can just grab off of the last time the VAP module made a call. Placing that calls ID here and hitting save will now cause this HTTP

07:52

get request to fire off. And we quickly see that it was a success.

We now have access to all of this data from our call, including the summary, which as you recall is the result of that prompt we added into VP to summarize the call for a sales team. We also have access to the analysis, which tells us if the call was successful.

This is perfect since

08:08

we'll use both pieces later in the scenario to update the lead's air table record with their call summary and interest level. So now that we're monitoring the call results, let's start implementing those next steps.

First, we want to determine whether the call was answered. So we'll set up a router to create paths for both answered and

08:24

unanswered calls. When someone answers the call and engages with your voice assistant, it's a valuable interaction that your sales team needs to know about.

Let's set up a filter to check if someone answered the call. On the first route, we'll add a filter with a condition that checks for calls where the data do the ended reason

08:41

equals customer ended call. Among all the data points available, this is our most reliable indicator that the call was answered and didn't go to voicemail.

On the second route, we'll do the opposite and set the condition to filter for cases where data ended reason does not equal customer ended call.

08:59

On the answered route, we want to know if the lead was interested in being sent a proposal or not. So we'll add another router that splits into two paths based on the call analysis, specifically the success evaluation.

Remember earlier when we told Varie to evaluate if the call was a success based on whether the

09:14

lead wanted a proposal or not. That evaluation is contained in this data here.

Since this data is simply true or false, we'll check whether the analysis is equal to true success or false failure on each route. For both routes, we want to update the lead's record with the call summary and note whether they

09:31

were interested in a proposal. So on each route we'll add an air table update a record module.

On the interested route we'll target the lead contacts based by this lead's ID and update the contacted on field two. Now add the call summary and check yes on interested.

09:50

We'll do the same for the not interested route but check no for interested. Now, our sales rep can instantly see when the lead was contacted and what transpired during the call.

The system automatically documents each interaction, just as a human rep would

10:06

take notes after speaking with the prospect. By the way, in case you're curious, this process technically performs a put request under the hood where it puts or updates something about the record without replacing it all together.

Of course, not all of our VP calls will be answered. This happens, people are busy or they screen unknown

10:23

numbers. That's why we build in a backup.

On the notwans answered route, we'll keep the steps from our initial build where we send that lead an email to schedule a call and we let our sales team know about them. With this all locked in place, let's run the scenario to make sure it's working.

Hey, it's Ben from Edge AI. Is now a good time to

10:40

quickly talk about your business needs? Yes, it is.

Great. Thank you.

I'd love to understand a bit more what led you to reach out to us. Are there specific processes or tasks you're looking to automate?

Yeah, looking to uh implement AI agents.

10:55

That's exciting. Implementing AI agents can really streamline operations.

Um are there particular tasks or workflows at your company that you find especially timeconuming or manual which you're hoping to improve with automation? Yeah, just hoping to um have the agents trained on the learning material so that

11:10

they can help my students. That makes a lot of sense.

Training AI agents to assist your students could really enhance their learning experience. Thanks for sharing that.

Would it be helpful if I sent a quick proposal your way showing how we could support you in developing those AI agents? Yeah, that would be perfect.

11:27

Excellent. I'll send the proposal to the email we have on file.

Thanks again. We've done a lot.

Let's zoom out for a second to recap. With our current scenario, we now have a system that detects qualified leads, places an AI powered voice call with no manual

11:43

effort, waits, listens, and reacts depending on whether someone answered and whether they were interested or not, logging the outcome accordingly. As of now, the voice assistant works, but it's not as capable as it could be.

Without knowing our lead's name or anything about their company and needs, the agents ability to pitch is pretty

11:59

limited. Our next step is to enhance our scenario by researching our lead with open AI and passing those insights to our Vappy Assistant, allowing it to personalize each call.

Essentially, we're going to have a a search feature that is going to not only just use the voice agent for uh doing the research

12:15

and getting more information, we're going to use OpenAI's uh search models. So, when people fill out that form we made, then we're going to use OpenAI to research the internet for that lead and get some information on that.

Then, we're going to send a call to them again using Vappy. This time it's going to be personalized with the information that

12:31

we got from that web search. So it's really, hey, we know this about you, but what else?

We're looking for this information on you. You can really get a a a complete picture of who this person is before we even booked them in for a call with our sales team.

So with the VIP module we're currently using, unfortunately, we're not able to feed anything into it, at least not at the

12:47

time of this recording based on the VIP module's current setup, but we can solve for this by switching out the VP module with a more custom approach using an HTTP module. So, let's drag it out of the workflow and unlink it.

And since we'll be borrowing some of the values from it soon

13:02

inside a new HTTP make a request module. Since we're making a manual request out to the Vappy API, we need to specify the URL just like we did in our existing HTTP module.

So, we can go ahead and copy that URL from the get module. Since

13:18

we're simply placing a Vappy call and not retrieving an existing one, we won't need the calls ID. So, we'll leave that off.

And we'll change the method type to post. The header remains the same,

13:33

but we need to add a second item where the name is content type. Value is application/json.

Content type is like labeling the envelope you're sending. Application/json means inside this request, the data is structured like a JSON object.

It's like writing English

13:50

or Spanish on the outside of a letter so the recipient knows what language to expect when reading it. Continuing down the module, we'll then set body type to raw, which means we're manually writing out the data we want to send to VP.

And we set content type to application/json,

14:06

which again tells VPY to expect data formatted as JSON. And I know many of you don't know what JSON is, but it's less intimidating than you might think.

It's really just a pair, the key and the value. Just like in a spreadsheet where you have the keys like phones, name, and

14:21

email. Then you have the values, the actual data that goes with each key.

It's an easy to read way to structure and share data. In JSON, everything sits inside curly brackets.

This is called an object. Both the keys and values need to be in quotes with commas separating each

14:36

key value pair. Don't worry if you're not a JSON expert.

Many free online tools can help you check if your JSON formatting is correct. So back in make down in the body of our HTTP call out to Vappy, we're going to add some JSON to save some time.

We'll grab some values from the VPY module we were using

14:52

earlier, including the assistant ID, which we'll paste into the JSON. And we'll also grab the VPY phone number ID off that old VP module and paste reuse it in our new HTTP module.

We're just doing it a bit more manually. The main difference here is in this

15:08

assistant override section. As it sounds, we are overriding the assistant within custom variables.

These are essentially placeholders that will be replaced by the lead's name, company, and the research we perform for each of them. This way, when we say, "Hey, Vapy, call this lead." We're also saying, "And

15:24

here's info about them to use on the call." Now, that we're going to be sending these variables into the assistant, we need to head back over to the Vappy dashboard and tell our assistant Ben to be expecting that information and instruct them on how to use it. In the first message, we can add the first name variable so our assistant

15:40

can greet our lead by name. In the updated prompt, we'll inform Ben that he will receive custom data via variables such as the lead's first name, company, and company research.

And to use these details to personalize everything on the call with the goal of pitching more effectively. With our assistant ready to

15:55

receive all that custom info, we just need to perform research on our lead. And for that, we'll be using OpenAI.

Let's add an Open AI. Create a chat completion module in line just before we make the Vappy call.

You'll need to set up the connection with your OpenAI account and make sure get an OpenAI API

16:10

key and also have some credits to use for this. To do this, you can log into your account at platform.opai.com login and click on API keys.

Then create a new one, making sure to copy and save it in a safe place. Then you'll click over to billing and add some credits,

16:27

making sure you have an active credit card set up. With that connection set up, we'll select the model.

Since we're doing research, I'm choosing GPT40 mini search preview. By the time you watch this, there may be other options.

The important thing is to choose a model that can do search. In the messages

16:42

section, we'll set the role to user, which just means this message is coming from a user, you in this case. The text content is where we place the prompt.

Essentially, we're telling it to serve as a research assistant that uses our leads information to generate a summary of how an AI automation agency could

16:58

help them with their needs. Let's see how it works by running this module only giving it an example company name and notes for testing purposes.

As you can see, it's running and performing research for us. But notice how the result is formatted in

17:14

paragraphs. While that might look fine, it actually causes a problem because when we try passing this result into our HTTP module, it needs to follow strict formatting rules.

Those paragraph breaks can quietly break things behind the scenes and cause the system to reject it since it won't be valid JSON. There are

17:30

a couple ways to fix this. We could either demand that chat GPT gives us our summary in JSON, or we could add a text passer in between the GPT and HTTP modules, which removes those line breaks for us.

Since there may be other situations when you need to transform data in your future workflows, it's

17:46

helpful to get some practice passing text. So, let's add the text passer replace tool.

For the pattern, we'll add this regular expression. This essentially finds any line breaks, so we can replace them with a new value, which we'll leave blank.

In other words, instead of the line break, it will be replaced by nothing, aka removed. We

18:03

want global match to be set to yes, so that it finds all line breaks, not just the first one, then no, for the rest of the options. Finally, we'll select the text we need to pass the result from open AI's research.

With this reformatting in place, we can now reference that clean text inside the

18:19

company research field of our HTTP body. Finally, we'll say yes to pass the response and close out this module.

Notice how we still have this red error icon on the second HTTP module. It's broken because it's trying to reference the call ID from the original VP module

18:35

that no longer exists because we deleted that module. Instead, it needs to reference the call ID from our new HTTP module, which manually makes that VP call.

Since we haven't yet run that new VP replacement module, we'll run it in isolation, feeding it a phone number and

18:50

whatever info just to get it to run. Once it runs, we can now access and reference the ID field from it for future runroughs of this workflow.

We're now ready for the big payoff. Make sure you add a new lead to your Air Table with a valid company it can research.

19:06

Let's run the scenario to make sure everything is working properly. Hey Mark, it's Ben from Edjai.

Is now a good time to talk about your business needs? Yeah, sure is.

Awesome. I'm following up on the form

19:21

you submitted. We build custom AI agents and automations for companies like View Mastery.

Anything from customerf facing bots to uh internal workflow optimization. I'd love to understand a bit more about what led you to reach out.

What sort of processes or tasks are

19:37

you looking to automate? I'm looking to implement AI agents to facilitate the learning journey on my e-learning platform.

That sounds like a great initiative. Enhancing the learning journey for your students can really make a difference.

Are there any specific bottlenecks or manual tasks

19:52

you're hoping to reduce with these AI agents? Just hoping to increase the effectiveness of our content by supplementing it with these AI agents.

Got it. So, you're looking to enhance the effectiveness of your content by providing real-time support to to your

20:08

students. That's a smart approach, especially with the extensive library you have at View Mastery.

Thanks for sharing that. Would it be helpful if I sent a quick proposal your way showing how we could support your goals?

Yeah, that would be perfect. Perfect.

I'll send that to the email we

20:23

have on file. Thanks again, Mark.

All right. Thanks.

Exciting. Our scenario is working and our voice assistant is now empowered to perform dynamic pitches based on our lead's unique information.

We're almost there. Finally, within build 3, we take this about as far as we can within a beginner tutorial like this.

Um, and this is a really, really

20:40

powerful thing once you've added in these extra features. So, after the call again, we're going to analyze that and deal with if they answered or didn't answer.

And long story short, if they answered and they said that they were interested and said, "Hey, yes, can you please send me a proposal?" Then we're going to take all of this research that we've done and all the information that we got from the phone call in order to

20:56

generate them a custom proposal and saying hey look this is what we want to kick off with you because you need to make a proposal in order to start any kind of services or most kinds of services but we can automate the generation of a proposal which can take hours and hours and hours for businesses and we can use all of this information we collected and they've said yes hey

21:12

can you send it over and then we use an application called Panda do and we can create a template of a document for our agency in this case and it's going to use AI in this case we're going to use a chatbt uh node here on make and it's going to take all of this information and write a personalized proposal on how

21:27

we would kick things off with them of what we're proposing in terms of the scope of work for them like we will do this this it sounds like you need this we can do this this is roughly how much it's going to cost etc and then using panda do we can send that as a e-ign link so that they're ready to sign and we get notifications about if they've viewed it if they've signed it etc and

21:43

so by the time you've done this we've automated everything from the initial point of contact where the leader said that they're interested in our services to learning more about them to determining if they're qualified for our offer uh to sending them a custom proposal and ultimately for them signing the dock through Panda do and they're

21:58

ready to kick things off with us. So that is an explanation of of build 3 and what we're really trying to go here.

I hope this been helpful to clarify things for you um because this is really really powerful if you can wrap your head around it. So please stick with it.

Once a lead has expressed interest, it's

22:14

the perfect moment to harness that momentum and transform it into something concrete, a tailored business proposal. Why wait for someone on your team to do this manually when we already have all the context we need?

With the help of OpenAI and Panda do, we can generate, send, and log a custom proposal without

22:30

anyone lifting a finger. So, in the final section of this course, we'll be tacking on a proposal generator to the end of our workflow.

We'll use OpenAI to create the custom text to plug into a proposal generator. We're going to be using Panda Do as our tool for creating and sending proposals.

22:46

It allows you to create templates with placeholders that can be filled in dynamically from your automation workflows. Here's how to create and configure your template in Panda Do.

Log into the Panda dashboard. Just create an account if you don't yet have one.

Then go to templates and click plus template. For sake of ease, we can select an

23:03

existing template to remix such as one of these business proposal or advertising sales proposal templates. I'm going to use a template I already created here.

Within a Panda Do template, you are able to drop in tokens, which are basically placeholders that can be replaced with actual values,

23:18

such as your client's company name. In this case, you can set up this template however you'd like, but as you can see, I've set mine up like this with a client introduction section addressing them by name.

In the goals and plan section, I've left room to insert information about my lead's goals and the services

23:34

I'll recommend and a plan for how I'll implement things. so you can see how it works.

We'll create those as variables. Over here in the sidebar, we'll add proposal.goals will be a paragraph or two summarizing the client's top priorities.

Proposal.services

23:51

will be a bulleted list of the recommended services. Proposal.implementation will be a concise execution plan to deliver the above.

In the pricing section, I've already added placeholders here. proposal pricing and for a breakdown of services and costs and

24:07

proposal.total, the full estimated project total. Down in the agreement section, we're requesting signature and including the leads info.

Once everything is in place, name it something relevant and save it because you'll soon be using this template inside your makes panda dooc

24:23

module. Optionally, you could spend some time styling this template with a logo, brand colors, etc.

Remember, this document will be client-f facing, so make it look and feel as professional as the service you're offering. With your Panda do template ready and tokenized correctly with placeholders, we're ready

24:38

to include it in the final sequence of our make scenario. We'll add the Panda Do create a document module, set up a connection with our Panda Do account, name the document based on the company we're sending it to, and select the proposal template we created earlier.

We'll fill in these values, giving the

24:53

module the lead's email to send this to and include all of the necessary info about our client, like the company name and the client's first and last name. For all these proposal tokens, we'll be generating these values with AI in a moment.

For now, we'll scroll down and say yes

25:10

to send a document because we want to email this Panda doc to our lead. Fill in the subject line using their name and company, write a short message, then hit save.

Next, we'll use OpenAI, create a chat completion module to help us write a clear, convincing proposal.

25:26

We'll select a quick and efficient GPT model, and in the prompt, we feed it instructions about its role as a sales expert with context about the services our mock company offers like AI automation and agent-based systems, information about the client, including the company research we did earlier in

25:42

the workflow, and the summary of the call our voice assistant had with them. Then we clarify its task to identify the most relevant services we can offer them and we demand the output to be in JSON format so we can make easy use of it in the modules after this.

If we go ahead and run this module only

25:59

passing in some dummy data for the call summary and company research. We'll see it efficiently generates this relevant info organized as JSON like we requested.

We're almost ready to plug these values into the Panda doc module,

26:16

but we first need to add a pass JSON module to prep the message content output from OpenAI, breaking it up into discrete values that are available as variables we can inject into our proposal document in the next step. To plug these proposal values into their slots in the Panda do module, we first

26:32

need the JSON module to run. While we could run our entire scenario, there's a quicker way.

go into our OpenAI module, grab the expected JSON format, and paste it as dummy data into the JSON module to test it. Perfect.

Now, we can see it's

26:47

passing correctly, and we can now go into our Panda Do module and plug those variables into their corresponding token slots. Finally, we want to bring things full circle by updating the leads air table record so that our sales team can see when the proposal was sent.

27:08

So in the proposal sent on field, we'll add a timestamp with now. This keeps everything tidy.

So when someone on your team looks at a lead, they know a lead was created, they were qualified, they were called, including a call summary, they were interested, a proposal was sent. No one has to chase a status

27:24

update. It's all there in one glance automatically.

Let's do a final run of this scenario to make sure all of the moving pieces are working correctly. Hey Brad, it's Ben from Edge AI.

Is now a good time to quickly talk about your business needs? Yeah, it is.

Great.

27:41

Thanks. I came across some info about View Mastery.

It sounds like you're really focused on helping developers learn VDJs efficiently. We build custom AI agents and automations for companies like like yours.

Anything from customerf facing bots to internal workflow optimization. I'd love to understand a

27:57

bit more about what led you to reach out. Are there specific processes or tasks you're looking to automate?

Yeah, I'm looking to implement AI agents for learning who are trained on the content library of my learning platform. Uh, that makes a lot of sense.

Creating

28:13

AI agents that can assist learners by leveraging your content library sounds like a great way to enhance engagement. Are there particular bottlenecks or manual tasks you're hoping to reduce with these AI agents?

Yeah, just hoping for uh the students to be able to comprehend and apply the concepts more effectively with the agents.

28:30

That's a smart approach. Empowering students to better understand and apply concepts with AI support can really boost their learning experience.

Thanks for sharing that. Would it be helpful if I sent a quick proposal your way showing how we could support new mastery with

28:47

tailored AI solutions? Yes, please go ahead and do that.

Perfect. I'll send the proposal to the email we have on file.

Thanks again, Brad. Thank you.

You're welcome, Brad. Thanks for your time today.

If you have any

29:02

questions later or want to discuss further, just reach out. Have a great day.

The workflow seemed to have worked as planned. Inside Air Table, we can see that the proposal sent on date was added.

Now, let's check the email that we told Panda do to send the proposal to. Heading into that email, we can see

29:18

that it worked perfectly. Our proposal dock arrived which the client can open up to go view their proposal which we can verify has been properly drafted detailing the client's specific needs and our plans to help them.

From here the client can sign the document, date it and finish it out.

29:34

An added benefit of Pandanda is that we don't need to create additional automation steps in make.com to track when leads view or sign proposals. Panda handles this automatically by sending notification emails to our email whenever a lead views or completes a proposal.

With this final stretch,

29:50

you've created a full system that not only identifies and qualifies leads, it translates interest into action. Of course, there are many other steps you could add to this scenario.

For example, the delay we added after the call takes place works in most cases, but if the call exceeds 5 minutes, the system would

30:06

incorrectly mark it as not answered, even if it was successful. The foolproof solution would be to implement a web hook that listens for the call to end, but that's outside the scope of this beginner build.

You could even add a whole extension to this workflow where you detect when a lead signs the

30:21

proposal and then onboard them with an automated client orientation workflow. I encourage you to get creative and add on to it to learn and challenge yourself.

For now, I want to share some troubleshooting tips to keep in mind as you go off on your own and build your own workflows. [Music]

30:38

As you take your next steps and start building your own AI powered automations, I want to be completely transparent. You will encounter issues.

platforms will change, tutorials may become outdated, and unexpected problems arise. This isn't a flaw in the system.

It's a natural part of working in a

30:54

rapidly evolving field. The truth is, even experienced developers spend a significant portion of their time troubleshooting.

I can't count how many times I found myself yelling at my screen because some seemingly simple thing wouldn't work. It's part of the process for everyone.

There's an old

31:09

saying, give a man a fish and you feed him for a day. Teach a man to fish and you feed him for a lifetime.

This sums up our approach to technical education. If you are spoonfed every solution, you'll end up just copying existing systems and won't be prepared for real

31:24

world challenges and you won't develop ways to differentiate yourself either. So, think of technical problem solving as a muscle.

Right now, it might be underdeveloped and using it feels uncomfortable. That's normal.

But with consistent exercise, tackling problems, finding solutions, learning from

31:40

mistakes, this muscle will grow stronger. It's a gradual process that develops through experience.

Each problem you solve builds your expertise, making future challenges less daunting. While you can't rush this journey, you can fully embrace it.

When you feel that sense of being out of your depth, try to

31:56

recognize it not as failure, but as growth. You're in uncharted waters.

You're pushing your limits and expanding your capabilities. The most valuable learning happens precisely in these moments of struggle.

So with that growth mindset, let's cover some tools you can use for troubleshooting. You probably

32:12

won't be surprised that I encourage you to leverage AI as you develop your skills. If you run into errors in your workflow by heading over to chat GPT or a similar LLM, you can lean on its ability to support you in solving the issues you encounter.

The trick here is to describe your issue in detail. Remember to give enough context so the

32:28

AI can grasp what you're building, the specific step you're on, and the expected outcome versus the problem you're experiencing. Include screenshots, error codes, and let it search for the latest information.

There will be times your issues are a little more complicated. For complex

32:45

problems, you can even try the deep research mode to get comprehensive information on specific platforms or issues. Sometimes the first suggested solution won't be the right one.

But as you engage in back and forth dialogue, as you see me doing here, where I am giving it more info,

33:01

the AI gains more clarity about your issue and can better narrow down potential solutions. And if you truly want to learn, don't just do what it says without curiosity.

Ask it to help you understand why the issue happened and why the new solution

33:17

ultimately worked. If you are hitting a wall, it's possible that others have already hit that same wall and climbed over it.

So, by doing a traditional Google search, you can look for existing threads or community posts about similar issues. Many problems you'll encounter have already been solved by others.

33:32

course, YouTube is another great resource. Search for recent tutorials that show updated interfaces and workflows or that highlight key details that could help resolve your issue.

We now have access to tools like Google AI Studio that can watch your screen and provide real-time support as you work.

33:48

Think of this as an experienced guide peering over your shoulder and helping you step by step. This eliminates the need to manually type out your problem and provide screenshots to chat GPT since you can give the tool permission to observe what's on your screen and guides you through solutions using voice assistance.

Take a look at how powerful

34:05

this is as it helps diagnose what's causing this error. I'm getting an error on this HTTP module where I'm trying to make a post request out to the VP API.

Any idea where I should start looking for a solution to this error? I see the workflow you've created and that you are receiving an error when trying to make a

34:21

post request to the VAPI API. Are there any error messages that you can share?

It's a 400 bad request error. Do you see the information here in red?

Okay, I see that it's a 400 bad request error. The

34:37

message indicates there's an issue with the JSON formatting of your request body. It seems like there's an unexpected character, either an expected or after array element in JSON at position 1167, line 12, column 51.

Would you like me to help you review the JSON

34:53

structure in your request to identify any syntax errors or missing commas or brackets? Yes.

Can you notice anything in here that is wrong? Looking at the JSON, it seems like the error might be in the company research field where you have an array containing

35:09

a single element and that element seems to be an expression. If you're trying to pass a single string there, try removing the curly bracket.

Also, can you double check that all your keys and values are properly formatted as strings? Learning to read and apply documentation is a critical skill.

Each platform has

35:26

detailed guides that are regularly updated. Often, you'll find information there that can help solve your issue.

If it feels overly technical, you can circle back to asking something like chat GPT to help you understand the documentation and how it applies to your specific issue. Finally, there are

35:41

plenty of online communities like the ones me and my team run related to the platforms you're using. Join Discord servers or other forums, search for similar issues, or even ask questions directly.

You'll be surprised by how helpful and supportive these communities can be. If your goal is to build real automations for real clients, things

35:57

won't always go according to plan. platforms will change and requirements will shift.

The ability to adapt, troubleshoot, and find solutions is what separates successful builders from those who give up early. The most successful people in this field aren't those who never encounter problems.

They're the ones who persist through difficulties,

36:14

who see obstacles as puzzles to be solved rather than roadblocks to progress. So, when you hit that inevitable moment where something isn't working, and you feel stuck, remember this is normal.

Take a breath. Step away if needed.

Then come back and work through your troubleshooting toolkit.

36:29

Each time you solve a problem, you're not just fixing that specific issue. You're becoming better at solving all future problems.

This resilience and problem solving ability will be just as valuable to your success as any technical skill you learn. So, embrace the journey of building your technical

36:44

skills. In the next and final section of this course, I'll show you how to sell your newly built systems to real world clients.

Now that you understand how AI automations work and can build them for yourself, let's talk about actually making money with these skills. But

37:01

first, let me destroy a huge misconception that many people have. You don't need to build the next chat GBT or create some revolutionary AI startup to make some money in the AI space.

The real opportunity is much simpler. Helping businesses to understand and implement AI automations like what you've just learned.

This is how I

37:17

monetize my AI automation skills, and it has been the most explosive growth I've ever experienced in my career. The good news is, if you've made it this far in the video, you're much closer to being able to tap into this starving market for AI automation services than you think.

But don't take my word for it. Let's hear from some of the world's most famous and successful businessmen.

If I

37:34

was 25 years old today in 2024, what would I do? What's a good sector to get involved in?

What business would I get involved in? I think everything is looking at AI now in a different way.

And I think AI growth is going to be exponential. So, anything to do with AI.

Now, what could that be? in the simplest

37:50

form is helping people use the technology. There's going to be a massive amount of people wanting to use it that don't know how to and they're willing to pay to solve that painoint.

So, is that consulting? Not really.

It's implementation and execution and so

38:06

helping a business do that transfer into a world where they're controlling their data and getting information from it. Now, the majority of businesses in America, for example, are between 5 and 500 employees.

So, they're small businesses. They create 62% of the jobs.

38:21

They want to use AI. You should help them solve for that and they'll pay you.

Mark Cuban is saying the same thing as well. That the biggest opportunity right now is helping small to mediumsiz businesses who don't understand AI and automation yet but desperately need it to survive.

And they're right. Here's why.

According to recent data, there are

38:38

1.7 million businesses in the US alone, making between $500,000 and $10 million per year. These are small businesses which make up 62% of all jobs in the USA.

These businesses know that they need AI automation to stay competitive, but they don't have the time to learn it

38:53

themselves. And there's basically no one there to help them at all.

All of the big consulting firms are looking at other big businesses, and no one serving this small business market who still need AI automations just as much as anyone else. So basically, almost all small businesses are starving for AI automation services like education

39:10

services in order to help them to understand what AI and automation is and why they need them. Then there's consulting services which help them to identify where AI automation can help them the most.

And finally, there is actual implementation services to help them build and maintain their AI

39:25

automation systems. Right now, based on data I've collected in my community, for every person offering AI automation services, there are over 1,100 businesses in the USA alone that need help.

1 to 1,100. The market is completely untapped and will be for a very long time.

And that's where you

39:41

come in, which is helping these hardworking small business owners to understand and implement AI automations so that they can keep up and survive through this AI revolution just like you. We've seen this exact same pattern happen when the internet came out.

The companies that help businesses to adapt to the web made fortunes as the

39:56

companies that they helped. I personally spotted this opportunity in early 2023 and started Morningside AI, my AI automation agency.

And since then, I've generated over $5 million selling AI products and services. And as a company, we're only just getting started.

And the best part is that, as we've proved in

40:11

this video already, that you don't need to be a technical genius to understand AI and even to build your own AI automations. You just need to be one step ahead of the businesses that you're going to help.

Let me show you the three specific ways that you can start making money with your AI automation skills.

40:27

So, there are basically three services that you can provide to help businesses with AI automation. First, there's education.

This is teaching businesses about AI automation, running your own workshops or presentations, or training their teams and creating courses. Businesses are desperate for someone who

40:42

can explain this stuff in simple terms to you, just like I've done for you in this video, what the hell AI is, what automation is, and what it can do for them. And after watching this video and a few others that I've made on this channel, you'll know more than enough to start educating businesses and helping them to sort of move from where you were at the start of this video to where

40:58

you're going to be at the end of it. I'll be covering which videos those are in a moment.

And secondly, you have consulting services. And this is where you analyze a business's operations and show them where AI automations can help them save time or make extra money.

You're essentially being their AI automation strategist. For example, you could recommend a lead qualification

41:14

workflow like the one you just made to help a struggling sales department. And third is implementation.

This is where you actually build and deploy AI automation solutions for businesses. Or better yet, like my agency, you can do all three of these.

You can do education, consulting, and implementation. But you don't have to do

41:30

it all at once. It took us like over two years to get here.

So there's no rush. And believe it or not, there are people with only a few months of experience in the AI automation space selling all of these kinds of services right now.

Education, consulting, and implementation. And the demand from businesses is increasing insanely fast right now because they're all switching

41:46

on and realizing they need to do something about this right now. But here's the thing.

You still have one problem, which is that you still don't really know enough. You're close, but you're not quite there.

The way to make money in the AI automation space or with any services really is to create what's called a knowledge gap between yourself

42:03

and the people that you're trying to help. Your knowledge gap is your money maker.

Businesses will pay you in proportion to how much more you know about AI automations workflows and their business applications than they do. So now while this video has taught you a lot, your knowledge gap is still fairly

42:18

small. But we can fix that.

Let me break down exactly what you need to do to extend your knowledge gap so that you can start monetizing this skill. This video is step one.

As long as you've been taking notes and followed all of the tutorials by building the automations alongside me, you're already

42:34

far ahead of most people who have no idea about how to build AI automations. Step two is building even more experience with AI automation so that you are more familiar with the platforms and better understand different ways that they are being used to help businesses.

I've only given you a taster here. This is a foundational knowledge

42:49

is the point of this video. But to do this and extend your knowledge gap, you can take my free course on school where you'll build another 5 to 10 automations.

Um, and the link to join will be in the description below this video. And this will further expand your knowledge gap without paying a dime.

Once you've done that, you'll have what I call a foundational knowledge. So you

43:05

understand the core AI concepts. You can build basic solutions yourself.

And you know what's possible for businesses right now. And then comes the big decision.

Do you want to go deeper technically or do you want to start monetizing what you already know? As we've already covered, building and implementing AI automations is only one

43:20

of the services that you can sell. Naturally, the technical skills needed to make money in implementation, actually building AI automation systems are far greater than just having a foundation, you know, so you need to be a lot more skilled in terms of experience and being hands-on with the

43:36

technology to deliver high-quality services for your clients. However, with a good foundation, you're basically ready to start having a crack at selling AI automation, education, or consulting.

So this decision of whether you go to education and consulting or actually building it depends on what you're really interested in. I'll use myself as

43:52

an example. I've always loved making things from like block houses when I was a kid to like brewing beer with my my granddad to tinkering with engines.

So when I hit this foundational level in early 2023 with my skills, I naturally kind of just dove deeper into the technical side. I kept building more and

44:08

more complex AI automations which led me to starting Morningside AI where we now build AI solutions for clients. But here's the thing.

A lot of people aren't like me. And chances are you aren't either.

They don't get much of a buzz out of the building side of things. And because of this, many of you going to be better at teaching and actually working

44:24

with people to help them understand and get value out of this technology than actually building stuff. But that doesn't mean that you can't make money in the space for you.

Using the foundational knowledge that you'll have after you finish that free course to actually sell AI, automation, educational consulting makes way more sense. So the key is being honest about

44:40

your strengths and your interest and setting real expectations on how technical you want to get in your career by picking whether you want to sell educational consulting or development services. You have a hard stop on how much you need to learn before you start taking action.

So this essentially prevents you from getting stuck in an

44:56

endless learning phase. I see it all the time where people just keep, oh, I don't know enough.

I need to keep learning and they're forcing themselves to do something they're not really good at or they don't want to do and they're procrastinating when they could be out there making money. So, in summary, your options from here are if you love building and want to learn more, then just keep going.

Keep following that

45:12

interest and that energy. Watch my free course tutorials in the description on my free school.

And then after you've completed all of those, go on to build your own projects and then ones for friends and family. Try solving your own problem or the people around you solving their problems with AI workflow automation.

And within 2 to 3 months of doing that, you'll be ready to start

45:27

trying to sell implementation to the people around you properly. However, if by now you haven't fallen in love with the building process, then it's probably best that you just finish off your foundation by doing the builds in my free course and then get started on monetizing.

Like, you don't need to keep forcing yourself to learn more and more and more and more. You already know

45:42

enough cuz you have a knowledge gap between yourself and your clients and you can start to help them by monetizing that knowledge gap. So, once you're clear on what type of AI automation business owner you are going to be, getting your first few clients is actually pretty straightforward.

And there are two main ways of doing this.

45:58

The first is through what are called warm connections. And this is by far the easiest way to start.

Instead of expecting strangers to trust you with your automation expertise, you can start with people who already know you and already trust you. What you're going to be doing is basically making a big list of all the people you know, all of the connections and friends that you have and acquaintances even, and reaching out

46:15

to them systematically to say, "Hey, I'm doing this AI automation thing. Would you be interested in having a chat to see how we could help you or your business?" And it's just consistently starting those conversations with these people in your network.

And then eventually, one of those doors is going to open and become your first client. I've covered this exact process many times here on the channel.

So on the

46:30

school post for this video, I'm going to add the all of my complete guides for warm outreach, including resources directly from my AAA accelerator program that will help you to fast track your first few clients. The second way is what I call the community content flywheel.

So this is how you build long-term momentum with getting clients for your business. Firstly, you need to

46:46

get inside my free community on school and then immediately start creating content about what you're learning. So this could be uh making YouTube tutorials about building automations.

It can be LinkedIn posts about workflow automation tips and tricks or whatever platform you prefer. But here's the key is that you need to share this content

47:01

back into the community. And we have over 160,000 members in my free school with the biggest AI business community in the world.

And that gives you an instant audience of people who are interested in the same things to help you get traction with your content as fast as possible. So let me give you a perfect example of this working in

47:17

action. We have a guy called Rory Ridges, a young guy from the UK, and he joined my free community and followed this exact process.

He took the free course in the community. He learned the basics and started posting simple tutorials on make.com and relevance AI and literally just started sharing the automation workflows that he had learned

47:32

from my videos. Like he wasn't trying to reinvent the wheel.

He was just saying, "Oh, I learned this cool thing. I'm going to make a video on it." But every time that he made a tutorial or a video, he would share it back into the community.

The community would watch it, give him feedback, and many of them would go and subscribe to his channel and become regular viewers. So, this not

47:47

only helped him to grow his channel faster, but it started to position him as an automation expert to his potential clients. So, now his YouTube channel brings him in enough leads to support his growing AI automation agency.

He's basically started the same flywheel that took me from zero subscribers and 0 with AI to on track to making $10 million

48:04

this year and over 500,000 subscribers in just 2 years. So, essentially, the community gives you an audience.

The content gives you credibility and together they bring you clients consistently. So, on my free school, there's going to be a post for this video, and I'm going to leave all of the links to my complete guide for creating content and generating leads, just like Rory and I have done.

I've done a video

48:20

on it, ton of resources. You can find that all on the free school.

What's really important to notice with both of these methods is that they start with giving value first. Whether it's helping your warm connections to understand and implement AI automation for free or sharing your workflow

48:36

automation knowledge for free through content, you have to start giving before you get. Now, I know all this businessy stuff may feel a little bit overwhelming or out of reach for some of you, but you will seriously be amazed at what baby steps add up to in this AI automation space.

You've already taken the first

48:51

step by watching and following through on this video. So, congratulations.

All you need to do now is keep this momentum going. And the next step for all of you is pretty clear now.

You need to jump in my free community. Go in there and drop an introduction post.

Let everyone know who you are, what you'd like to do with your AI automation career, and then

49:06

start working through my free course material. It's there for a reason.

And I've poured everything that I've learned about AI automations and building AI automation businesses into videos like these and they're all in a nice sequence for you to work through on school each time you complete a video. You can click the little check box and keep stacking

49:21

those small wins and keep that momentum going until you get to where you want to go. And all of the resources I've mentioned in this last chapter, the selling part of this video will be on a school post for this video within school if you go to the YouTube resources tab and then it should be right there.

And don't forget to check all of those resources out there. And of course, if

49:37

you've made it this far, could you please do me a big favor and just leave a like on this video? You can drop a comment below.

Tell me what you like the most or what you'd like to see next. And click the share button and send it to any loved ones or family or friends or anyone so that they can start learning these valuable automation skills for the future as well.

So, all of these actions

49:53

just help my videos like these reach more people in the YouTube algorithm and I'd really appreciate it cuz I put a lot of work into these for you. And of course, subscribe to the channel for more content like this, helping you understand AI automations, how to build them, and more importantly, how to build businesses and make money around this incredible opportunity that is AI.

50:09

you want to check out my 4-hour guide on how to build AI agents, I've got that up here. It's just like this one, but on AI agents.

Got a ton of great feedback from it. So, if you want to keep learning, that's a great place to go.

But aside from that, guys, thank you so much for watching. That's all for the video and I'll see you in the next

Summary

Transcript