🚀 Add to Chrome – It’s Free - YouTube Summarizer
Category: AI Tools
00:00
In this video, I'm going to walk you through eight of the best use cases for OpenAI's new ChatgBT agent mode. These use cases save me hours doing repetitive tasks in my day-to-day and in my business.
So, we're going to break it down and look at agents that can handle everything from lead generation, helping
00:16
you with the accounting all the way to ordering what you're going to have for dinner. But more on that later.
Without further ado, let's jump into it. So, first things first, we're going to open up Chat GBT.
And in a new chat, I'm going to under here under tools, activate agent mode. And now let's take
00:32
a look at our very first use case. And here I want to highlight probably one of the biggest uh like let's say new capabilities that agents provide.
And that is the fact that agents can create entire presentations. So what I'm going to do is I'm going to copy a very simple prompt.
create a presentation with
00:48
everything someone needs to know about the new chatbt agent mode from OpenAI include pricing benchmarks as well as the top use cases for agent mode and I can just send this one off and we can see our agent gets started actually creates its own desktop its own
01:04
virtual computer that it can then use to a do the research as well as finally creating the presentation. So here our agent has opened up a browser and you can see it's identified the exact article it needs to open which is the chat GBT blog article.
01:20
It's then opening up the OpenAI website and what's happening here is it's actually reading what all the information about ChatGBT agents and what we can do is I can fast forward here. And we can see that our agent worked for 13 minutes, did four searches, looked at 38 sources to
01:36
identify everything it needed to identify about Chat GBT agents, and finally it created this full presentation. So I can open this one up and you can see this is a full seven slide presentation with uh graphics, with tables, with charts.
You can see
01:52
this is so much better than anything that AI was able to create previously. And the most impressive part is this is in PowerPoint format, meaning that I can simply just download this presentation and import it into either Google Slides or PowerPoint and continue to edit this.
02:08
Now, what we'll see is in the prompt, I didn't give Chat GBT any instructions around design. So, it came up with quite a generic design, but we can actually guide the design behavior of these presentations that our Chat GBT agents will create.
This time, I've tried that exact same prompt, but I also added some
02:24
additional instructions about the design of the slides. So, I've added here a little visual style guide.
I've said the font should be Inter. I've added the different brand colors that we like to use at 9x.
Gave it some instructions about the type of illustrations it should use and also attached our logo
02:39
that it can include in the presentation. Here we can see it worked a little bit longer.
It took 25 minutes, but we now have a presentation that follows our exact style guide and uses our brand colors. So you have the ability based on how you prompt this agent to create
02:56
different styles of presentation. Now right now I've provided this information in the prompt.
What you could easily do is create a chat GBT project and provide this information or even save it as an overall instruction to your Chat GBT account. Meaning whenever you create
03:11
presentations, it will follow this exact style guide. And just in case you're already wondering, a little later in the video, I'll show you exactly how you can access all of the prompts that I'm using for the eight different use cases so that you can try them out for yourself.
Now, the next chatbt agent use case is
03:28
this data analyst agent. And another reason why a lot of people are talking about chatbt agents is their ability to work with and handle spreadsheets.
So for this use case, I've attached a spreadsheet which contains some student progress and this is real data from one
03:44
of the companies we are working with. I've obviously anonymized both the names and the company name, but this is real progress from students in terms of the different courses they are taking um in our different online AI training.
So per student, we can see for each of the online courses what they've viewed, how
04:01
much they've completed, and their progress. And I'm simply giving this data to the agent along with quite a detailed prompt.
So here I'm saying attached as an export of student progress for employees of Acme Inc. who are taking our AI and automation online
04:16
training program. Now I'm providing some very specific context in terms of the order in which the courses should be taken since in that data it only refers to the course name and doesn't give any specific order.
And then I'm also um laying out the different training
04:31
programs that we have. So we have our AI operator training program which contains these courses and we have our AI builder training program which contains these courses.
And finally then I'm giving the agent instructions that I want it to generate an interactive Excel dashboard
04:48
as well as a student progress summary presentation that we can share with the HR department or the L &D team of this company. So I just need to activate agent mode and I'm going to fire this one off.
And jumping forward, we can see this one ran for 8 minutes. Please find
05:04
below the interactive Excel dashboard and the student progress summary presentation. It's already giving me some key highlights from the data.
So, first of all, we have this Excel. And now, this looks simply like the Excel that I shared.
But if I open this one up, we'll see that Chat Tubet has
05:20
created some new tabs. So we have here the course metrics which is now listing per course what is the average completion rate, how many students have completed it, how many are in progress and how many have not started and we also have some graphs to go along with it.
And we also have an employee
05:36
progress. So now listing out per employee from Acme where each of the students stand and again with an accompanying chart.
Even more insane is the presentation that chatbt has now created for the L & D team of Acme Inc. We get a little overview about where
05:52
their employees stand that there's total 19 employees taking our courses with an average completion rate of 43%. We have a different charts on which courses are they completing also which of the programs the AI operator or the AI builder where the different students
06:08
stand and we even have a leaderboard in terms of which of the employees are really taking our course and maybe which of them needs a little gentle reminder from their manager as well as some recommendations and next steps. Obviously, this is something that I would QA.
I would check that the numbers
06:24
are accurate, but this is actually running in Python. So, it's not a large language model actually running these numbers.
It creates a Python script that will crunch these numbers, create the Excel file, and then use that as a basis for this presentation. Now, this next
06:40
use case is a lead research agent or lead generation agent. And this is going to be particularly powerful if you're working for or running a B2B business.
What you need to ask yourself, and you do need to get a little bit creative here, is think about what signals are potential customers giving off that
06:57
could identify them as a valuable prospect for us to get in contact with. Now, since at 9x, we offer AI and automation training and enablement.
One signal that we identified that we wanted to investigate were companies that are hiring AI transformation roles. So, what
07:13
I've instructed my chat GBT lead generation agent, you're an expert lead generation agent specializing in finding leads for B2B companies. I want you to identify at least 50 companies that either currently have or in the last 3 months had open job positions around AI
07:30
transformation or enablement. any jobs where the role involves driving AI adoption rather than working on AI products and the company should either be based in US or in Europe since that's where most of the companies that we work with are based and you can see our agent
07:45
is getting to work going through different job portals. It's had a look at LinkedIn.
It's going through the top uh job portals and you can see here it's identified some software engineers and it's probably going to reject those since it doesn't match the condition. So, I'm going to fast forward to the
08:01
end. And we can see here our agent is finished.
It worked for a total of 44 minutes. It looked through performed 61 searches and looked across over 400 different sources.
This would have taken me way longer. And we can see here in a very neat little table.
We have our list
08:16
of 50 leads. For each of those leads, we have the company name.
We also have where they're based as well as the role that they're currently hiring for and evidence of AI adoption enablement role responsibilities inside the role and we
08:33
have the different sources. So in one click I can simply open up the source and verify here they are looking for a senior director of AI transformation.
This is a company that we should maybe get in contact with. Now, building such an agent in any other platform would
08:49
take a lot longer. And all of this was done in one single prompt.
Again, this is something that I could set to run on a schedule. So, maybe every week I'm bringing in the next batch of leads and you can test many different signals.
All you need to do is update the prompt. So,
09:05
this next chat GBT agent use case solves the number one question that I get asked the most, and that is, how do I keep on top of the latest in AI with how fast things are moving? So here we're setting up an AI news watching agent or a news monitoring agent.
So here's a very quite
09:22
a detailed prompt saying you're an AI model daily watcher agent. Here I'm specifying its goal.
So once every day at 8 a.m. it should discover any news articles, press releases, product updates around openai chat GBT, Anthropical Claude or Gemini.
But here
09:38
you could basically replace this with whatever news you are interested in. I've given it some more guidelines around what it should treat as a significant news update, the rules in terms of that it should use its built-in search feature.
So for this in the sources, I'm making sure that web search is enabled. And finally specifying an
09:56
output. So if it finds some significant news items, it should basically spit out a list saying the AI model updates with the current day and then one news item under the other.
If no items are found, simply say no significant updates in the
10:11
past 24 hours. So, I'm going to fire this agent off and we'll see the first results.
We can see our agent is working through. It's basically identified some news articles and now it's reading for instance here about Google's Gemini's new deep think feature which was written
10:28
on August the 4th which is the current day now Gemini 2.5 deep think. So there's a few articles around this new deep think feature.
All right. All right, so our news agent has run.
You can see it worked for 12 minutes and completed 20 searches and read through 139 sources. It read through a bunch of
10:46
article on Gemini, on Chat Gibbt, on Anthropic. What is the latest news?
You can see it's actually opening up websites. Now, out of all of those 139 sources, instead of just regurgitating all of that information to me, it decided there's that I there's only two
11:02
bits of news that I should really be interested in, and that is an update on Anthropic blocking OpenAI's API access and Google launching their new Deep Think feature for $250 a month. Now, obviously, this is something that just ran once.
For me to
11:17
turn this into my daily agent, what I want to do is I'm going to click on this little schedule button. What's going to happen now?
It's going to suggest a prompt. In most cases, you want to disregard this because you can see that it's actually just regurgitated its answer as the instructions.
So, instead,
11:33
what I'm going to do is I'm going to copy my initial prompt because that has all the information that my agent needs to run every single day. I'm going to hit schedule again, paste that initial prompt in here.
This is going to be my AI model daily news watcher agent. I'm going to set it to run daily at 8:00
11:49
a.m. and hit save.
And now this is basically taken care of for me. Every morning I will receive a notification on my phone at 8:00 a.m.
And in one click I have all the latest updates in the world of AI. And you can adapt this to whatever news you want to follow.
To
12:06
give you an idea about what this looks like when you have this scheduled. So here was one that I had set up yesterday.
And you can see I received this notification on my phone about the AI monitoring task. There is some updates.
In one click, I can simply open that notification and I see what are the
12:23
top five things in the world of AI that I need to know about. You can see our agent worked for 9 minutes, read through hundreds of different sources, performed multiple searches, and gives me exactly what I need to know directly in my phone.
And this is something I receive now every single morning. and a keeps me
12:40
informed a lot more than I previously was and saves me a bunch of time. Now, real quick before we get into the next use case, if you want to test out any of these use cases for yourself, in the description below, you'll find a link to our website where we have these free AI and automation tutorials.
And we're
12:56
going to be creating tutorials for every single use case so you can easily copy any of the prompts I've used today. All right, on to the next one.
Now, this next agent is an invoice tracking agent. If you're anything like me, I am terrible at keeping on top of my invoices.
And this often gets me in
13:13
trouble with my co-founder PIV, who takes care of our accounting. And now, while some platforms will send us emails whenever we get an invoice, and this is something that can be totally automated, there are always those platforms where you don't get an email and you need to actively log into the platform to
13:30
download your invoices. So, here I've set up an invoice tracking agent.
What I've done, I've got this running on a schedule. So here I'm giving the example of a cleaning service that we use and I need to download the invoices.
So I've told I've basically instructed it for
13:47
once a week to go in log into the platform and download any invoices and keep track of them in a spreadsheet. We can see here the agent is smart enough to identify the login screen.
It's even clicking on the login button here. It then sees that there's a drop down.
This
14:03
whole page is in German. The agent is able to understand everything.
Now we're getting redirected to the login screen. And here the agent is going to ask me to take over.
So I can simply here click login. Now I'm going to be given access to the agent's virtual computer.
And I
14:19
can just go ahead and now type in my email address here or in my case I'm going to log in with Google. Now be I would always advise be quite careful about what access you are giving to these agents.
Uh even OpenAI recommend uh exercising very strict caution. In my
14:36
case, I'm happy to test this out uh for you guys all to see. So, I'm just going to log in with uh Google.
Once logged in, we can see our agent is able to identify which parts of the page it needs to click on. So, here it clicks on which is the German word for invoices.
14:53
It then loads the invoices page and it's going to go through and download every single invoice that we have and add them to the spreadsheet. And the final outcome is we have a spreadsheet with all of the information my co-founder needs for accounting and I have direct
15:10
links where I can open each of the invoices. This is literally something that is saving me hours from my day-to-day admin.
So this next chat GBT agent use case is all about staying up to date with your competitors and reacting quickly whenever they make
15:25
changes. So here I want you to imagine that we are competitors of the AI automation platform Zappia.
Now the AI and automation space is more competitive than ever. So we want to be alerted whenever Zapia make any pricing changes so that we can react quickly.
To set up
15:41
this use case and set up this agent, the first thing we want to set up is a Google sheet. So, I've got a simple Google sheet set up here called zapia.com price monitoring.
And here I've simply listed the different plans that they have on their pricing page and how much they cost. With that all done,
15:56
we can set up a very simple agent. So, here it's goal is to monitor zapia.com's pricing page for any changes and alert me when their plan prices differ from the latest values stored in my Google sheet.
I'm providing it with baseline data, the existing Google sheet. And
16:12
here I'm basically um giving the exact name of the Google sheet I want it to reference. I have um then the source to check which is the zappy.com pricing page.
This is the exact URL. And now here I'm telling the agent its workflow.
So first of all I want it to open the
16:27
Google sheet and I'm telling it that it should do this via the API using the Google Drive connector. So what I have here I've got agent mode enabled and in terms of the sources I've also toggled on Google Drive.
So, I've given the chat GBT agent access to my Google Drive and it will be able to view that
16:44
spreadsheet. After it checks the latest uh information from a spreadsheet, it's going to go and visit the Zapia pricing page and compare the difference.
So, let's take a look at the first run that this did. So, here we can see our agent used the Google Drive connector to read from Google Drive and it successfully
17:00
picked up that zapia.com price monitoring Google sheet. It then went ahead and opened up the pricing page of Zapia here.
It's in reading mode and it was able to read all of the information on the page and you can see it gives back its response that there were no changes. The prices listed on Zapia's pricing page match those stored in the
17:17
Google sheet. Now, this is something that I just set up a schedule to run every week and inform me of any changes.
Now, to test out if this one's working, what I can simply do is I'm just going to go into my Google sheet and say that the pro plan is $29.99. That's the
17:32
latest information that we have. Then you can imagine this is the next day the agent is running again.
Again it's reading the information from the Google drive now with the updated information. It's then going to again visit the up-to-ate information from the Zapia page.
And we can see now this time it is
17:49
actually given us an alert and showing that the pro plan the old price what it saw in the Google sheet is $29.99 and now as it's saying on the website is $19.99. So you could basically set up this agent is so simple.
If you were trying to do this in any other platform, you would have to take care of the logic
18:06
of scraping the page, storing the information here with just one prompt, one Google sheet. You can set up a agent that monitors your competitors and updates you whenever anything changes.
Now, our next chatbt agent use case is a
18:21
competitive analysis agent. And here we're leveraging deep research.
So, you can think about this use case can apply to any sort of deep research analysis that you need to do. and you can adjust your agent accordingly.
So in our case, we've asked our agent to deliver an end-to-end competitor analysis for our
18:37
industry, which is the B2B AI and automation training industry. And it should follow this eight-step process in terms of how a good competitor analysis should be done.
And finally, we have the deliverables that the agent should give, which includes an executive brief with
18:54
the key risks, opportunities, and next steps in this industry. slide deck, 10 to 15 slides, source workbook, as well as a raw benchmark CSV.
So when it com any data that it collects or any sources that it finds, we can refer back to later. And finally, an email handoff.
19:12
Now, if you're wondering how I came up with that prompt, I didn't. I actually used Chat GBT for this, and this is something I always recommend if you are not an expert in the field.
I'm not an expert in competition analysis. So what I did, I simply asked the 03 model, which is one of the best for these
19:28
reasoning tasks, what is a competitor analysis? And it gave a very detailed answer about what a competitor analysis is.
And then I asked it to list out a full list of instructions on how a competitor analysis should be performed. And here you can see it provided an end
19:45
to-end playbook that we can adapt to pretty much any market. So here is its the basically the steps that I'll eventually be using in the prompt.
And after it completed that I asked it is it familiar with the new chat GBT agent mode and it was and then finally I
20:00
simply asked it to can you please prepare a prompt that I can use to prompt the chat GBT agent following all of those best practices that you just found. And we can see here now chatgbt03 has created the perfect prompt that we can use for our agent and it's basically
20:17
made it dynamic. You can adapt this to whatever industry you need.
So going back now to the result of that competitive analysis agent, we can see here this worked for 25 minutes, did 15 searches, and read through 136 sources. Of course, starting with the
20:32
deliverables, we have this executive brief, nicely formatted, the competitive landscape, the insight and applications and the recommendations. We have a full slide deck presentations which cover everything from strategic recommendations, what is the current market overview and growth, where is the
20:49
competitive landscape. Unfortunately, 9x isn't appearing here yet.
So maybe if you subscribe to the channel and get a bit more noise in a few months time will also be here when I prompt chat GBT to perform such an analysis. We can see price versus benefit.
a bunch of
21:04
different analysis that we would probably have paid uh a lot of money for and is at least a very very good starting point. We also have as we asked for a source workbook.
So all the sources that was used as well as any data sets that were benchmarked and
21:20
going all the way to the bottom we have this email handoff so that I could basically share this with my co-founders. I think this is a little bit too long, but you can see where we're getting at where this level of detail would have been so hard to get to with just a simple prompt without using this agent mode.
And I would really
21:36
recommend you think whenever you need to do such an in-depth analysis where maybe you're not an expert in the field, give it to AI first and see what results you get as a starting point. Now for the eighth and final chatbt agent use case, this is not so much workrelated but a
21:53
bit more on the personal side. Sometimes it's my turn to do the shopping and I've been working late and I'm not going to make it in time.
And rather than my girlfriend getting angry at me, ChatGBT is going to come to the rescue. So, I'm going to have agent mode enabled.
And now I'm simply actually going to send it
22:10
a voice message. Hey, Chat GBT, can you please order on Flink um all the ingredients that I would need to cook spaghetti bolognese?
Please try and uh optimize towards organic ingredients. Um and don't worry
22:27
about the staples like olive oil and salt and pepper. I already have them at home.
So here is my prompt to chatbt. Flink for anyone wondering is a food delivery app here in Berlin.
So I'm just going to send this one off and my agent is going to set up its desktop. So the
22:45
first thing the agent's going to do is identify that it needs to open up the Flink browser. And so it knows exactly which address to go to.
And here we go. So we have Flink loaded.
This is normally a mobile app, but you does have a desktop version. So the agent's just going to select shop now.
And so while
23:01
this agent is running, I can continue with the current work that I'm doing or the current video that I'm recording. And now it's asking me, could you please enter your delivery address?
I'm just going to click take over. And I've entered my address.
I'll say I'm located here. And now I can say that
23:18
this is an office since I'm getting it sent to my office. And I can then hit save.
Now that my address has been saved, the agent knows exactly where I should. I'm going to select finish controlling and hand this back over to my agent.
As I said, now I can continue
23:34
with whatever work I'm doing. So the agent has identified the first important ingredient, which is spaghetti.
So it's searched for that in the search bar. It's added the item to the organic.
It pointed out out of all products, it found the organic uh spa spaghetti, in
23:50
this case, spaettini, and it's added that to my cart. So, checking back in with our agent.
It's been working for 18 minutes, and you can see it's completely filled the cart here on my Flink account with everything I need for my spaghetti bolognese. I'm also going to show some clips about how it decided in the
24:06
process to choose some of the organic products over the non-organic ones. But you can see here everything I need to cook my dinner tonight.
Let me know in the comments below if you think it got it right in terms of what should be in a traditional spaghetti bolognese. I know that can sometimes be contentious, but
24:22
basically I was able to keep working and now chatbt has ordered my dinner for me. So I just showed you some use cases that you can use for yourself.
But something else you can do with the AI agents that you build is actually sell them for profit or use them to get you promoted. And if you click right here, you can
24:39
check out a full multi-hour course to learn exactly how to build and sell AI agents. Everything from idea to build to actually implementing these in real businesses.
This is one of the most unique opportunities we've ever seen. So, I'll see you over there.
Otherwise, until next time, happy automating.