8 Insane ChatGPT Work Use Cases! (automate anything) transcript

In this video, I'm going to walk you through eight of the best use cases for OpenAI's brand new ChatGPT work. It is an AI agent that lives on your desktop. It can see the files you're already working on, has access to its very own browser, and can take action inside the tools you use each and every day. So, instead of just giving you an answer, it actually goes and does the work. Every one of these use cases are saving me hours every single week.

So, we're going to break it down and look at agents that can handle everything from building your reporting dashboards, researching your entire lead list, all the way to kicking off a job from your computer, and then finishing it from your phone. But, more on that later. Let's jump straight in. All right, so I have ChatGPT open. Now, very importantly, it doesn't matter what system you're working on, you want to be using the dedicated ChatGPT desktop app for this, and not the web version, because some of the use cases that I'm going to show you will only work when you use the desktop app.

So, let me just make the app full screen. So, if you download this for the first time and open it up, this is what you should be landing on. Make sure you have ChatGPT selected here in the top left, and not CodeX. And then, what you want to do is this toggle at the top here, change it from chat to work, and now you are in this new ChatGPT work mode. And for our first use case, I want to start with something that probably blew me away the most when testing this new tool, and that's creating interactive dashboards.

So, what I want to do here is build an actual dashboard that I need for myself and my team to keep track of how our YouTube videos are performing. So, the very first thing that I'm actually going to do is head over to Google and try and get some inspiration. So, I'm just going to search for YouTube Analytics dashboard. And if we take a look at the images, so we can see a few example ones here. This one already looks a little bit interesting.

So, let me actually open this one up, this image here of the dashboard. So, checking out this image, it looks pretty good. We've got the key data at the top, some individual graphs. And then, all I'm going to do is simply copy this image, and then paste it here into ChatGPT work, and that's now attached. And then, I'm providing a very simple prompt. So, recreate this YouTube dashboard from the screenshot attached as a live site, which I can share with my team, pull the live numbers from our YouTube channel, and also from BigQuery, which is our data warehouse where additional YouTube data is kept.

Now, before I send this prompt off, I'm going to add a couple of things. First of all, a custom skill that I've created called YouTube channel. This basically teaches ChatGPT Work how to connect to our YouTube channel and pull in data because there isn't any inbuilt plugin for that, but there are a couple of other inbuilt plugins that we're going to be using. So, here on plugins, the very first thing I'm going to do is add the data analytics plugin, which teaches it how to connect to our data warehouse and also what type of queries it should run, and also the sites plugin, which is going to use to build the interactive dashboard.

So, let me send this one off. And if we take a look at the result of what ChatGPT Work has created. So, here we have our live dashboard ready to view. Let me open this one up in our browser. And there we go. So, we have the YouTube channel overview. It's obviously pulled in the logo from our YouTube channel, which is very cool. We have the key stats at the top here. And this looks pretty much one-for-one from the screenshot that I sent it.

Very, very simple. We have all the key metrics from the last 30 days at the top here. A few different breakdowns of the daily views, the average view duration, and the total watch time. And then we also have a per day view. So, which days are actually performing the best for us in terms of views on YouTube. And maybe this is something that we can then use to customize when we release our videos. And then we also have a full table down here featuring our latest videos.

Now, what I want to point out here, it didn't necessarily nail this from the very first attempt. So, you can see here, I was giving it additional feedback after its first version. So, something I didn't like about the first version is that when I hovered over these graphs that it wasn't actually showing the numbers like it is now. It wasn't as interactive as I wanted it. And all I needed to do was give it that feedback.

So, here I just asked, "Is it also possible to make it so that when I hover over the charts, it shows the numbers for the the day or the specific segment?" It worked for another 4 minutes and made that update. Now, if I compare this to the old way of doing things, I used to work with BI and data analyst teams quite a lot. They often have a big backlog from all departments within a company, and typically you put in a request like this, and if you're lucky, you get it back in 2 or 3 weeks.

Now, we get a first version up in a matter of minutes, and you can also see how fast it is to iterate. That example that I just showed there of making it more interactive, 4 minutes later, it was live and ready to go. And with these interactive dashboards, they make it very, very easy to share this with your team. I can just click share, choose do I want to make this available to everyone in my organization, or only share it with individual people.

Uh onto our next use case, and that's using ChatGPT Work as your very own web automation agent. So, inside ChatGPT Work on the desktop app, it comes with its very own browser that it's able to use. So, at the top right of the desktop app, you can see that we can toggle the side panel, and then access the browser. Now, one task that I've actually been meaning to do, over on the 9 x blog, we can see we use a very consistent style for our different blog article feature images.

So, they're all illustrations with similar colors that we use. And what I want to do is basically download a bunch of these, so I can actually train up ChatGPT Work to create images like this, which we'll do a little bit later on in the video. But for now, for me to actually collect all of these images together, I don't want to go through and actually right-click and save all of these. Let's be really lazy here. I'm going to copy this link over and give that to ChatGPT Work, and then I've asked it to visit our blog article page, download the first 20 feature images of the different articles that we have, save them all inside a folder.

I also want it to create a CSV, which has that blog article title and the image, so that the AI will be able to later learn between what image corresponds to what article. And here we go. We can see it's confirmed it's going to use its browser skill to inspect the live blog listing. And I can already see in the browser on the right-hand side, it has opened up a new tab. Let's click on that one so we can see what the AI is doing.

Yep, so it's gone over to our blog. I'm not controlling this anymore. I'll just make it a little bit bigger for us to see. You can see it has its very own mouse, and unlike a person that would actually like right-click, um which it actually can do, it will probably here write itself a quick JavaScript to download all of the images. And we can see after just a few minutes, it is finished with the task, downloaded and verified all 20 feature images.

It's giving me the folder where they are all stored as well as that CSV file. Let's actually close this browser also here directly in ChatGPT Work. Again, on the right-hand side, I could open up files, and here we go. There we see inside this folder that it created all 20 of the feature images that it has downloaded and that CSV file, which basically tells it, "Okay, this was the blog article title, and this is the corresponding image." Now, this was obviously a very simple example of web automation, but where this comes in particularly handy is if you are working with a tool that doesn't have a dedicated API, a way for AI to connect to it using a connector or an MCP server, you can have ChatGPT Work run on a regular schedule, let's say once a month when you know the invoices will be coming in, to log into that platform on your behalf and download the invoice and put it wherever you need it to go.

All right, onto our next use case, and this is something that absolutely nobody wants to be doing anymore in their work, and that's building out Excel models. And for this example, I thought I'd throw quite the challenge at ChatGPT Work, and that was having it come up with a full cohort analysis, something that is quite complex to do in Excel and would take a lot of time if you were doing this manually. So, what I prepared, I have five different exports from different tools that we use, which each contain important data if you are running a cohort analysis.

So, we have some stats from our Circle community in terms of what members we have, what is their activity log, also from the platform that run our live events, importantly our Stripe transactions to know which members are paid and which ones are free, and then even an export from an old tool that we were using before we migrated to the new community. So, here I gave ChatGPT work quite a detailed prompt laying out the five exports that it has access to in its working folder, and then I asked it to build me a proper cohort analysis as a single Excel workbook.

Now, very importantly, in this prompt, I made it very clear to ChatGPT work that it needs to be using formulas in the whole Excel workbook. I don't want it to be putting hard-coded numbers where I can't actually trace whether it was accurate or whether it made a mistake. Everything needs to be an actual formula inside Excel. And lastly, what I added to the prompt before firing it off was the inbuilt spreadsheets plugin.

Now, amazingly, while I went off and did something else, ChatGPT worked for over an hour bringing this Excel file together, creating the full model. You can see here, it worked for 1 hour, 7 minutes, and 44 seconds. We can see if we open this drop-down all of the steps that ChatGPT work did during this time. And importantly, here I want to point out that it recognized straight away the exports that I gave it. It was quite a substantial amount of data, so around 35 MB.

There would be no chance if you tried uploading this to the web version of ChatGPT that this would work. It probably would say, "Hey, there's too much data. I can't work with it." And that is the beauty of using the desktop app. It worked through literally thousands of rows to come up with this analysis. So, let's go and take a look at that final Excel file that it created. So, here I can just open it up in Excel. And wow, here we go.

Here is our full cohort analysis. You can see all of the tabs of the raw data that it basically collated together. We have the people, the active month, and then more importantly, our cohort analysis here in the retention tab. Let me make this one a little bit bigger so it's easier to see. And what I really, really like is, as I said, all of these cells, we can see it is not hard-coded data. It is a formula that we can actually verify.

For instance, if it has also decided to do a cohort analysis the way that we maybe wouldn't do it. We could specify it to change the formulas. We also have this same graph split up by source, so we can see, okay, what was the cohort coming from the different channels where we attract customers, whether that's YouTube, LinkedIn, ads, or organic. And we have separated cohort analysis underneath here. And let me check out here as well.

Wow, the findings tab, complete with graphs that you can actually edit in Excel. This is not just images without any data behind it. All of these are pulling in live data from the Excel model that it is created. And we have some recommendations here as well. Now, importantly, as I mentioned, this one used the inbuilt spreadsheets plugin. And what's great about it, when you type @spreadsheets, you can see you can also use templates.

So, there's a bunch of out-of-the-box templates. For instance, in this example, I just had it create an Excel file from scratch. But imagine at your company, obviously, companies that are already running, there's probably a bunch of Excel files that are floating around that already matching the exact formula that you want, the exact way that you like to do the analysis. You can actually create that here as a template, and then you're not using ChatGPT Work to actually come up with the model from scratch, but rather it will basically put your model to work.

So, you can just have it plug in, do the manual stuff that you don't like doing, have it plug in the raw numbers according to the exact model that you provide. All right, onto our next use case, and that's using ChatGPT Work as your very own lead generation or lead research agent. So, for this one I actually set up my very own lead research skill, where I taught ChatGPT exactly how to look for companies that may be interested in what we offer here at Nine X, which is training their teams on AI.

So, I'm going to use my lead research skill. Now, whenever I use this skill, I need to provide a few different things. So, I'm just going to paste in the conditions for this search. So, I need to basically provide what region that I want it to currently look for. So, for this one I want it to focus more on the European market. I'll do a separate search for the US. And then tell it what signals it should look for. So, here it should be currently hiring for AI automation or ops transformation roles, that they recently announced a new AI initiative, and that they're 50 to 500 employees in typically business operator heavy industries.

Then finally, I put in here how many companies it should try and find, and for this demo I'm just setting it to 25. And most importantly, what it also has access to inside our lead research project folder is I have a full list of companies that we are already working with or have worked with in the past, so it knows to exclude this from the search. Now, before I send this one off to give it the best chance to succeed, I'm actually going to upgrade the model's effort here.

So, instead of just having it high, I'm going to set this to ultra. This does consume more usage, but you'll see why this is important in a second. So, let's send this one off. All right, now lead research agent has finished. It worked for over 20 minutes. So, we can see here, if I take a look what it actually did, it spun up different sub agents. We also see that here on the right hand side. So, for each of the different areas that it decided to look into, so it's created one sub agent for Berlin and Hamburg, created another sub agent to look into Munich, Vienna, and Zurich, and another one looking at the UK and Ireland.

And the beauty of these sub agents is that the work happens all at the same time. They all go off and do their research, and then basically report back to the main agent. You can see that it also created an additional sub agent to make sure we're correctly excluding any companies that we've already worked with. Let's take a look at the results. If I can open up the Excel file here directly inside Chat GPT work. And here we go.

So, we've got 25 AI training buying signals. I already see one company here that I know from Berlin, which is very cool, Coro. We can see here that Coro just opened up a principal AI transformation role to set company-wide AI targets, coach every team lead, and make AI the default way of working. Probably when they hire this person, we should get in contact and see if they may be interested in some of our training for their team.

Now, very importantly, for every signal that it's reporting that it actually found, it gives us the exact URL so that we can verify that information. So we see a lot of them here from companies that are actively hiring in the area of AI transformation. If we have a look here, another company, Peter Park, posted a team lead business automation and AI role to own company-wide AI strategy. This is coming from LinkedIn. Let's copy this link and see if we can verify this one.

And here we go, there it is, team lead business automation and AI from Peter Park that they're currently hiring for this role. Obviously, we now know that this company is focused on AI transformation and maybe they could be interested in any of our trainings. Now, nicely, I will blur these out here, but it also gives us a potential decision-maker name of who we should contact as well as their title. And you can see it's actually found quite high-up people, which always is very important whenever you're doing B2B selling that we have the chief operating officers here, the managing director, the COO, the founder and CEO.

So some really good contacts and even went ahead and found their LinkedIn profile so we can directly reach out and gives a summary of why they fit to NineX as well as the confidence rating. Now, if you're interested in testing out this lead research skill for yourself or any of the use cases that I'm showing in this video, in the link in the description below, you'll find a link to join our free community. There I will share all of the prompts that I use and any of the skills and plugins needed to make any of these use cases work.

Once you've joined the community, just head to the YouTube video section on the left-hand side here and find this exact video. Underneath all our videos, we share any resources to download. So you'll have all of the skills, all of the prompts, and any instructions that you need. All right, let's keep things rolling with our next use case, and that's keeping on top of your expenses. So for this one, what I've done is on my computer, created a dedicated folder called receipts.

So now, whenever I want to track an expense, all I simply need to do is take a photo of the receipt and then drop that one here into this folder. It doesn't matter if you want to do this for tracking your personal expenses or for your business, keeping on top of things and having things ready for your accountant. And over in Chat GPT Work, I've created a dedicated skill for processing these receipts. Now, normally this is set to run on a schedule, but here I'm just going to run it once manually so we can see what it does.

And if we keep an eye on our receipts folder, Now, it may have looked like that file has now disappeared, but what it's actually done is moved it into the processed folder. So, that's the way that this one is set up. Whenever there's receipts dropped here into the root folder, it should actually go and process them. And once it's done that, it's going to move them here into the processed folder and sort it based on the date of that receipt so we have everything organized come tax time or if we need to check our expenses.

And more importantly, it's actually also added that expense as a row here to our Excel file. Let's take a look at how it worked through this in Chat GPT Work. You can see the very first thing it did was view the image. So, correctly found that receipt image inside the folder, identified all of the key fields. And so, let's take a look at our receipt log and we can see here it successfully added on that expense to the bottom row, categorized it.

We have all of the details just from that one simple picture. Next use case, bulk editing data across tools. Now, the situation here at 9X is we get a large amount of our website traffic coming from the links that we place in our YouTube descriptions, especially if one of our videos, fingers crossed, takes off, then obviously people check out the descriptions and sometimes click on the links. Now, in most cases, whenever we're giving away a free resource in the video, that will always be the first link in the description.

But one new type of link that we've added to our latest video is this one here. If the viewer, maybe you yourself, happen to be working at a company interested in AI training, we want to make it as easy as possible for them to get in touch with us. So, we want to test out adding this as a link to our YouTube videos. Now, we have a large catalog of videos, I think over 130 at the moment. And for me to actually go through and update all of the description to add this link in would be a very painstaking task made even more complicated with the fact that every single link that we add here in our YouTube descriptions are basically short links so that we know exactly which video, so this has like tracking parameters behind them so that we know exactly whether someone visits our website, what video did they watch to bring them there.

And this is incredibly vital information for us. Basically, a few things I would need to do here to do this myself. First of all, I'd need to go into the tool that we use to create these custom tracked short links and create a dedicated one for each of our 130 videos. But then actually going into the YouTube dashboard and then manually going through and editing that in every single description, that is something that I want to avoid at all costs.

So, let's see if we can have ChatGPT work do this for us. Now, what I've already asked ChatGPT work to do, again using that skill that allows it to connect to our YouTube channel, is basically go through and list out all of the videos that we have set to public on our channel and then create a CSV with the ID of the video, the title, the date it was published, the amount of views, and then most importantly, create a markdown file that features the description.

And you can see here it's gone through and listed out all of the videos on our channel. And for each one, we have here the description that we can just preview. As you just seen for our latest video, we added this call to action for companies that are interested in team training. And in some cases, we actually have no, like let's say call to action above the fold at all. And in this case, I would just want to have immediately that team training link.

Now, this step of having ChatGPT work first go and actually download and save all of those descriptions is incredibly important for a couple of reasons. First of all, then it already knows the exact structure of our descriptions. But second of all, because in a moment I'm going to be asking it to actually go and live edit the descriptions in YouTube using its YouTube connector that I have set up. And of course, things can go wrong.

So, if anything does go wrong, we have all of the descriptions saved here that we can always revert back to. So, now I've prepared this very detailed prompt at the bottom here we can take a look. So, I'm telling it that I would like it to add the following call to action to all of our YouTube descriptions. And here I've given it the text and told it that there's going to be this short link that it's going to be the dynamic link that it needs to create.

Now, again we use a tool that doesn't have a dedicated connector for creating our short link. So, I've created a skill that has taught ChatGPT work how to connect to our tool short.io to come up with these custom links. Told it the exact tracking parameters that I want to have on every single link. And more importantly, it needs to take the video ID as a dynamic field. So, then we know exactly which video they watched when they visit our website.

And then I have a few different conditions. So, if the video description starts with a lead magnet link, so to download the resource of the video, then you should put this one underneath. If it doesn't start with a call to action, then just immediately put this as the first line. And otherwise, if it starts with a call to action for the user to subscribe, then actually just replace that subscription one, which is like this one where this is an example saying if you enjoyed this video then subscribe.

We just want to have directly here that training CTA. And now to test this out on a smaller set, I'm only going to ask it to do like I think what's this? Like the 20-30 videos that we have published this year. So, let me send this one off. All right. And ChatGPT work has finished with its first batch of update all the videos from this year. Let me actually get rid of this sidebar. So, here we can see them all listed out, which is really nice.

It's letting me know exactly what it changed. Um I have a link to the video where I can check out the update. And more importantly, it's also listed here so I can see actually it has gone ahead and created the short links dynamic for every single video. Let's take a look. So, this one had previously the subscription call to action. Let me open up this video. And perfect, there it is as the very first link in the description.

As I said, this would have taken me hours. In fact, I just would not have been able to to do this if it wasn't for Chat GPT Work. Of course, you could use like other automation tools, but the fact that it needs to understand the context of what's in the description, replace different characters depending on what's being said. This is just a perfect use case for a tool like this. Now, next use case using Chat GPT Work as your very own image generation agent.

Now, I know what you might be thinking, "Yann, we already know we can create images in Chat GPT, but by actually bringing this functionality into Chat GPT Work, you can actually create images inside of workflows you already need to do." Now, if you remember earlier on, just to recap, Chat GPT opened up our blog and downloaded the last 20 feature images so I could get a sense of exactly how these are created, the style that we use, and how these look like.

And now I've basically asked it to go further. So, saying I have three new blog articles that I need to create feature images for, and I have given them some titles. So, Chat GPT Work vs. Claude CoWork, What is Chat GPT Work and How to Set It Up, and Seven Browser Automations Every Marketer Should Run in Chat GPT Work. Can you please carefully analyze the format of the images that we currently use? You should get a good sense of how these images are created.

Make sure to also take note of the exact resolution. And then, for each of these new blog articles, I want you to come up with three suggestions cuz my experience in image generation, it might not nail it the first time. So, asking it to come up with several suggestions for each blog article means I'm going to have a much higher chance of finding one I like, and I can just pick that for the article. And we can see that this was not a problem at all for Chat GPT Work.

It worked for 12 minutes. Image generation does take a little bit longer, but while it was doing this, I could go and do something else. Here, it created nine feature image suggestions, all normalized to the standard format, and saved in this folder. Let's actually take a look at the folder where it saved the images. Let's go and take a look at some of these. And yeah, absolutely, it has nailed the style of the the way we were doing these for our blog articles.

So, this is the web automations that marketers should run in Claude CoWork. You can see here it's obviously always putting seven elements around a web browser, very, very nicely done. We have the Chat GPT Work versus Claude CoWork actually picking putting in the icons in our exact style. And then finally, the suggestions for what is Chat GPT Work. Now, what I really, really like here as well, you can see maybe if it's gone in the wrong direction, and if we want to improve it, all of the prompts that it used for actually generating these images without me even asking it, it actually saved it here in a markdown file.

So, we have here generation prompts. I can then go and carefully review these and see, okay, if it's gone in a direction that I don't like, we know exactly why. Now, obviously, this is not only limited to creating images for blog articles. If you're working with any sort of creative where you have already gotten some good results from AI-generated images, and also, of course, that they suit your industry. Maybe you're coming up with uh different meta ad variations.

Maybe you need some images to put in your emails. You can now bring this workflow in to Chat GPT Work, have it come up with detailed prompts, try several variations, and have it while you're doing something else, creating all these images for you in the background. All right, onto our final use case, and that's that you can run any of these agents from anywhere using the inbuilt remote feature. So, for this one, I'm actually going to also start recording the screen of my phone.

And here I am in the Chat GPT app on my phone. Now, very importantly, as I mentioned earlier, Chat GPT Work is also available in the cloud version, so on the web version, or also here on your phone, but that is not where you access this feature. So, in the left-hand menu here, we're going to head to the section called remote. If you don't see it, click the three dots where it says more, and we can see here it actually needs to use my Face ID to access it.

And now, what's amazing, in a second, this is going to load all of our Chat GPT work sessions that we kicked off from our computer. For example, this is the one where I had it rewrite all of those YouTube descriptions for all of our videos launched this year. I wanted to see this as a first test. Now, imagine that I'd kicked this one off while I was at my laptop, but now maybe I'm out and about, maybe having lunch, and I see, "Hey, this task is finished." Now, I don't need to wait to be back at my computer.

What I can actually do here, I'm just going to ask it, I'll even use a little annotation, "That's perfect. Can you also now run it for all videos from 2025?" And now, I'll send this one off. Now, what we've actually got, we are also recording at the same time my laptop, and you should see that that has actually come up and has started running. And I think this is really, really an amazing use case when working with agents.

These tasks typically take a lot longer than chat, so you can kick off a long-running task. As you saw in some of these use cases, some of them take up to an hour, and then you can always check in. Now, very important that because we're using the desktop version, it has access to our computer's files. Now, obviously, whatever computer you're working on needs to be on for this to work, but with this feature, you literally have the power of these agents wherever you are.

So, those were my top eight use cases of the new Chat GPT work. And if you're a Chat GPT user, I would urge you, stop using it simply as a place where you ask questions, and start using it as the thing that actually does the work for you. Let it do the browsing, let it reach into your actual tools, and build the final deliverables. So, look at your work week, find the thing that you redo the most often, and hand that one over first, and from next week, you'll be saving hours.

Now, as I mentioned, all of the skills, prompts, plugins that I used for these use cases are available in our free community. Link is in the description below. And if you happen to be a founder or CEO looking to get your team [music] upskilled on AI, as you've seen, there's a brand new link in the description below where you can book a call with us. Or otherwise, if you just want to keep learning right away, according to the algorithm, you should be watching this video next. >> [music]