10 Incredible Use Cases: ChatGPT Operator Agent (Must See!)

Look at this. Worked for 15 minutes. Today we'll be taking a closer look at OpenAI brand new product Operator. This is the very first agent that remote controls your browser that actually works. And I've tested all of these. Okay, this is obviously a very potent idea that hasn't found the right execution yet. And I'm not saying that Operator is the end all, be all product. It's a research preview. But OpenAI only showed off a few use cases that are very basic. Booking a table, booking a hotel. This thing can actually already do way more than that. And in this video, I'm going to show you what possibilities me and my team found inside of OpenAI's operator. Like researching multiple websites and putting all the data into a separate spreadsheet, or doing the same thing and creating a custom PowerPoint presentation, or transferring data from one source to another one, like a Google Sheet or Notion. And finally, we'll even look at a use case where Operator is using another AI to get results that it couldn't get anywhere else on the Internet. Before we dive into this one by one, I want to share two things. The first one is that I understand that this is inaccessible to most people as it is behind a $200 paywall. But that's why people like me are here that are going to test this out for you and show you what is possible. I think this product category is extremely promising. And this is the very first version of the product that I have tried where I feel like I will be using this regularly. There are several use cases from this video already that I saved to my tasks and that I will be using regularly moving forward. Is that worth the $200 spoiler for most people? No, probably not. Nevertheless, I do believe in this category. And that brings me to my final point before we get into the actual use cases, which is that I wanted to go further in this video than OpenAI did in its presentation. Nevertheless, I hope you find the use case that might make your life easier. And with that being said, let's get into the first use case that is possible today with OpenAI's operator that you might not have seen yet. All right, so one of the categories that a lot of people have been curious about is this use case of using Operator to research a specific topic. Now, whether that's going to different blog post posts and pulling the data together, or it's actually finding the websites that it should be looking at, this general use case of putting an AI agent in front of a keyboard and mouse to gather some data from The Internet is just something that a lot of people would find useful. So I came up with a few test cases to see how this will perform in practice. Let me start with the first one, which is kind of the simplest, and then we can kind of work our way up. So this one is based off the comment of Colin Bailey that is plain and simple. He said, just tell us how we can make money. And from all my research, I have not found a direct way to make money with this yet. I mean, I guess there's some indirect ways, but what I did is I told it to research five viable online business opportunities with low startup costs and then I also wanted the estimated startup costs with potential monthly revenue. And I just ran this and it worked for seven minutes right here. First up, it just found random articles of a Google search that asked for online business opportunities. Not the most sophisticated method to do do this I suppose, but that's what you can expect. Then it opened them up successfully, navigated all the way through the various cookies, like clicked away from the ads, things like that, scrolled through the whole thing as it scrolls. It uses its vision recognition, the one in GPT4.0, to capture the information on this site. And then although it did find two different articles to do this, it decided to just take the data from this one, which by the way was 17 use cases, and it presented it back to me in this format, which is five use cases. I mean it did the things that I wanted, startup costs, potential monthly revenue, but rather than looking at multiple sites, it just looked at one and rather than giving me all 17, it just gave me five of these. So it worked, but I don't know, that's 2 out of 10, maybe a 3 out of 10. On execution of this, I like the fact that it used LLM to produce some of this extra data, as this was obviously not a part of the article. So that's nice and shows a lot of potential here. But this simple query, I mean, not really what you want to see out of this. By the way, this was one of the prompts that I also ran through Anthropic's Prompt Improver to see if a more detailed prompt like this that tells it about the goals, the requirements, the action steps, really breaks it down into the various components, performs better. Now turns out that the result is pretty much the same without giving me monetary values on startup costs, potential revenue, etc. So it actually performed a little worse. One interesting thing that occurred here was something that I saw in multiple operations and I'm quite impressed with this, because it's really good at navigating some of these sites. In this case, it went to the entrepreneur blog, used the search function, and when it got to an error page, it managed to navigate back out of it and get back into searching successfully. And then again it pulled up multiple sources, but the results all come from one source that it doesn't even cite. So again, a lot of potential for improvement independently of the fact how well written your prompt is. Okay, so then I took a different approach because I realized that, hey, it did what I told it to, right? It researched a topic, it searched for it, it found the site. So here's the same task, but approached from a different angle. The initial prompt I gave it here is quite simple. Analyze and execute the following tasks. Scrape the data from 10 different blog articles on how to use OpenAI operator. So I wanted to research different use cases or articles or any details on operator that I might have missed. And this resulted in the longest working time that I found. Look at this. Worked for 15 minutes because here I told it to actively not just search for multiple articles, but also to scrape the data from them. So this seemed to have worked really well. Let's review the video of the entire operation here because the result that you can see here is pretty much a summary of what you would have got from the OpenAI blog post, but pulled from all of these different sources. So it opens up one source, goes to the Bing search, opens up the next link, opens up the next link, opens up the next link, and so on. I think you see the point that just opens up all these various sources and then when it looks through them, it pulls the data together and this is sort of where the task ended. Then I followed up with another prompt which basically said summarize all the use cases into a Google Doc. And then it just needed a quick Google Doc sign in from me and it went on to compile this Google document that I created by itself and pulled everything together. Now, sure, it's just the copy paste job of the results over here, but I really like the fact that it can do this reliably. It can work with Google Sheets, Google Docs, so spin up new ones. As you'll see in other examples, it can transfer data and things like that really quite reliably. And then if we review the video here, you'll even see that in the end it always goes to the share settings and it changes them to share this with anyone with the link, copies the link and successfully includes it. So this part really works for Me. And also this data that it's scraped from the 10 different sites is actually a mixture of the various results here, making this a little more advanced and specific use case. A real success in my eyes. And I mean, you could do this yourself, right? You could pull up 10 articles, you could copy everything, you could run it through LLM to summarize it, but that takes a lot of time. That is not just booking a table at a restaurant or booking a hotel or something rather simple like that. Heck, Even operator needed 15 minutes to complete this task. And if you can find two or three of these that you will run maybe not every day, maybe once a week, let me tell you, that's a lot of time you can save here. But nevertheless, seeing something like this for me is a serious use case. Why Operator might add to your productivity. Okay, I was actually ready to call it a day and give up on this use case and just accept the fact that the summaries will be this concise. But over the next 48 hours, we actually returned to this one because I thought it was particularly impressive how Operator navigated the Internet and then just pulled together all the data. It's just that the very limited length of the output was really hamstringing this. So we ran a session in our community trying this out, and together with some of the community members, we came up with this idea of actually telling it to save a summary after each article to a new Google Doc with a simple prompt like this. And it actually did it. Check it out. This is the result of the 15 minute long operation. It looked at five different articles and it saved a summary of each one to here. This is way better than summarizing five different blog posts into one paragraph. Now I have one paragraph per blog post. Much preferable. We tried to push this further, but sort of failed. I want to show you anyway, so we just did one thing which was we told it for every three screenshots, I want you to add a summary of the written content in that screenshot to a Google Doc. And it did this. Do this. But this was super messy. I would not recommend it. It did work. So this is a workaround on how to make lengthy summaries with Operator. But this was the document that created right there. And as you can see, every sentence just starts with deep seq R1, the topic that we're researching. And it's just, I don't know, it's just not that good. So obviously these are limitations that with time will be loosened up and there will be competing products but for now, I wouldn't recommend this workaround, but this one is really incredible. And then, as you might have noticed, there's a separate part to the prompt here. And I actually use this in multiple prompts now. And I want to dedicate a separate sect section to this. So let's talk about it. Because this command really allows it to do a task from A to Z without asking your permission multiple times. So in our community, I already wrote up a little guide on how to exactly do this step by step. But I can summarize that back here to you because it's quite simple. If you run an operation for the first time, you will encounter it asking for permission on multiple steps. So for example, on this summary of the different blog posts, where it's supposed to add a summary to Google Doc like, so it always asks, hey, I did one to find more articles, should I continue searching? And then you have to come back and manually say, yes. This defeats the whole purpose of this application and in other operations it asks multiple times, can I rename this image? Can I download this file? Can I run another search? So what do we found in an office hour session that I run in the community is that if you specifically tell it what not to ask you for, it won't do it anymore. So if you just generally say something like only ask for my confirmation if you're spending money, it will ignore that and it will ask you anyway. But if you say that and then follow up with if you need to find more articles or continue searching, move forward, then as you can see, this conversation becomes one prompt, one result, exactly what you want from it. So I started including this building block into multiple prompts. Now, so the general workflow is just you run through the manual process of telling operator that it is allowed to continue, and then you take these little snippets and reintegrate them into this building block that allows it to proceed independently of what it thinks. Now, obviously this is just one little prompting trick, and many of these are already appearing and will keep appearing, and that's really what the community is here for. On a video like this, I can show you what we've done, but I can't keep you up to date on every little trick that we find or that other community members find. So yeah, I just want to point out if you're into this type of stuff, if you want to get the most out of it, I'm going to make the community the best place on the Internet to learn how to use OpenAI operator. Just because I'M obsessed with this myself and I hope this little snippet helps helps you out in executing your own operations seamlessly rather than having to confirm every little action. Okay, let's move on to the next one. But something like this is definitely going to find a spot in my task that will run regularly because having a research assistant that googles all the newest articles on any given topic and then summarizes it back to you in a doc like this Maybe not worth $200, but in my books this is a real use case that I will be returning to regularly. Next up is a very simple one and a lot of people asked if it could handle files, so I just gave it this little test of uploading a picture of Keanu Reeves. I don't know, I had this laying around on my desktop as one does and I told it post this picture to the ChatGPT subreddit. Now it needed me to log in and then it asked if it should use a different subreddit. Again, I could use the prompt snippet here before and if I customize it and include some of the wording in here. So if I say feel free to post on a different subreddit if this one isn't available or automatically log in for me with my Google account and then you give it the credentials, it will not ask. These follow ups point here being is that it actually successfully navigated to the ChatGPT subreddit, created a brand new post after I logged in and encountered a barrier. You need some karma on Reddit to post on a subreddit. So then it moved away and went to the OpenAI subreddit and uploaded the picture there. Now it didn't go through with posting it because I don't want to be sharing pictures of Keanu Reeves on that subreddit. But as you can see, the whole point here was that it has this file manager and you could for example give it multiple images and it will share them across across the Internet. Now obviously this could be used to spam Facebook groups, LinkedIn, etc. But this is nothing that spammers haven't been able to automate with traditional tools like a webdriver. The difference is now this becomes a whole lot more accessible and you can move around digital files like this without much friction. You literally just upload it to a new conversation, tell it to post it to a specific site, you can even give it the URL and then you can save the task. And whenever you need to do that you can just let it do all the Clicking for you a million ways to use this. I just wanted to demonstrate that it can store and then use files for you if you add them as an attachment to a brand new conversation. Okay, so let's look at some data transfer use cases because this was not just very requested but also very interesting, right? Having it find the data in one place and then put it into your database or Excel sheet in another place. Obviously something that a lot of people have to do manually all the time. And in many of those use cases using something like traditional automation just doesn't make sense because setting up the automation would took like five hours to automate away a task that might take five minutes. But quickly prompting operator with the links to the Google spreadsheets that you need to work with is something that takes a few seconds and in return it saves you five minutes. And then every time you need to adjust something, it takes a few seconds to adjust something. Saves you multiple minutes that does line up again. Even experienced person doing this with code will take 5 or 10 times amount of time of just doing the damn thing manually. And this is where operator shines because it's so simple and it figures out all the details on its own. So this is the example, ask it to copy data from a Google sheet to another spreadsheet and then it asked me hey, do you have the links to the sheets? And I actually created two Google sheets for it and gave these to them. And in this interaction they did have a few follow up prompts as you will see in the video. So for example, it just asked me for permission to paste the data and then I followed up and wanted to create a visualization and a forecast for the next 10 years and to place the charts in a new tab. So what you can see on screen right now is the entire process and parts of this worked really well. Parts of this didn't work at all. So copy pasting the data, flawless. I actually ran a few more examples of this every time it came to copy pasting something from one source to another. It does really really well at these tasks. So I can see myself actually using that. By the way, this data, I literally just said name data 1, data 2, data 3. But as you can see, it created the chart successfully. But with the forecast when it came to that, it started struggling a little bit because I wasn't specific enough. I realized that now in hindsight I realized that I just told it create a forecast for the next 10 years, but I didn't tell it how I wanted to forecast it. So obviously it should forecast the amounts, but do I want it to create a separate column for the forecast or do I want it to extend the amounts like I didn't tell it? And one thing that I really learned with all my testing of this so far is Operator does not like to ask follow up questions like ever, unless it's about permissions or performing an action and it just wanting your approval of its actions, it will not ask, hey, should I do it in the same column or should I use a separate column? Something that a real person would probably ask. It just does things just like ChatGPT, which great in many cases, but in a case like this I'm not even sure what exactly it's doing at this point. Work with Excel formulas. I think you could get it to work if you really know what you're doing, but if you just blindly tell it things like hey, forecast this data or create some visualizations, it's going to try, but I think it needs a little more direction here from somebody who actually knows what technique would be used here in Excel to forecast. Or I guess you could use ChatGPT on the side and ask it what formula you could use and then give it a specific formula. Nevertheless, in the end I had to correct it and tell it, hey, the forecasting thing didn't really work and it just figured it out. It just went cell by cell and created all of this different data with a little hiccup here in the end. But what it did successfully is create two brand new tabs. One of them with a graph of the current data and one of them with a graph of the forecast, which I consider a success. It created the graphs, it transferred the data. Forecasting didn't work so well. Data transfer, something that you can absolutely use this for. Now here's another example of that and this one works between different applications. So in the AI advantage we use Notion a ton, basically the entire operating system for both the community and our content production engine for the YouTube. I built in notion. And right now we have six team members testing various Operator use cases inside of a Notion database called Operator Use Cases. Now Philip from the team actually went ahead and he used Operator to update the Operator use Cases database in our Notion automatically. Is that amazing? So he tested the use case of using Notion and then Operator entered all the details as you can see in the screen recording right now. So this works like a charm. And as you might be able to tell, I'm really excited about this because we have so many manual tasks in Notion with the community and products we Created, for example, with the very first course that we did two years ago, we put together a library of over a thousand prompts. And in Notion, we created card views for them. So it looked like a catalog, but each one of those needed a picture. So we needed to manually head on over to Midjourney, generate an image, then move it over to Notion. And there was no API for these card covers, so we just had to manually pick the midjourney pictures and upload them and repeat that a thousand times. With Operator, this would be really simple because actually this is a finding from another team member, Michal. He found that the best way to work with Notion databases is, is actually in the card view. It's really good at this button like layout. So for any of you Notion nerds out there like me, if you create card views for your databases, Operator is going to be able to interact with them most reliably. As you can see, her table works too, but card views work even better. These are just the little tips and tricks that we're finding on a daily basis and then resharing to the community and the YouTube here. So, look, some of this stuff is pretty powerful, but it's not that straightforward, right? You might have to include an extra line in your prompt, or you might need somebody like me in this case to make you aware of the fact that a specific thing is even possible with this. This really does feel like the early days of ChatGPT or GPT4, when it came out, where there were thousands of unexplored opportunities and a little prompt snippet could really influence the way you work. But I feel like here it's even more impactful because this thing actually does stuff doesn't just tell you how to do things. And what I decided to do is to ReDirect Our entire AI Advantage community in this direction. I've always been saying that these assistants are just a precursor to agents. And now that we have the first proper agentic product from one of the biggest players, we're doing something we're calling Operator February. So we basically pivoted most of our events, all of our guide production, all the community discussions, to cover Operator, and to help anybody who's a member of the community get the most out of this product. We have a dedicated space where you can discuss different use cases, and me and my team are sharing them just like I'm in this video in that space as we discover them. Plus, there's discussions amongst members erupting underneath, pinpointing different possibilities or routes that can be taken with an operator. I really believe that this is the next evolution of AI and that here it starts getting so useful that this product will eventually get so useful that the mainstream will not be able to ignore this anymore. Just like the ChatGPT moment or now the Deep Seq moment recently, this will have its moment too. The difference is this is not as intuitive as you need to know what it can do. What are the limitations? How do you get around those? And my goal will be to make the AI Advantage community the best place on the Internet to educate you about that. Obviously a lot of it is also going to trickle down to the YouTube channel just like this video. But I just wanted to let you know that if you share my vision for this and you want to get as good as possible ad operator, join me and all the other members in the AI Advantage community and let's explore the possibilities of new tools like this together. All right, that's my little community promo that I really wanted to do this because it's going to be all about Operator and similar products moving forward. And now let's look at the next AI use case that you can do inside of operator today. Okay, here's another one that I really liked and this is really impressive because it was a one shot prompt and it actually did all of these things in sequence. So the prompt says compile a competitor analysis for the top five subscription based e commerce platforms in the pet supplies industry. Compare their pricing, marketing strategies and customer reviews. Then create a concise 5 slide presentation in Google Slide summarizing your findings and share it with insert Gmail address. And let me tell you, this one worked like a charm. It found five different subscription based pet supplies companies. It looked at reviews and looked through their websites for all the information we asked for. And then it went to Google Slides to create a brand new PowerPoint presentation, so to say. And look at this thing operating away. I mean isn't this impressive? You can get drafts of presentations on any topic in the click of a button if you have the right prompt preset. This goes way beyond the power level of even the best prompt inside of chatgpt. I mean, come on, this is amazing. There's no denying that like this is where all the arguments like oh, it can only book a table or a flight. I can do that myself. Go out of the window like come on, five slide presentation from one prompt. Not bad. We're just getting started here. So yeah, this works well. And this is another one of those presets that I'll just have in my tasks and whenever I need new research on something, this is a great way to do that. Just research a niche and create a presentation from it. And literally one click and here's another one like that. It's really good at these research tasks if you give it some source to store the data. So this one is also from Philip and we'll put all the prompts here in the description below. But basically he told it he was looking for affordable pocket camera and he wanted product link pros and cons. And the pricing. Basically doing some rudimentary product link research. And as you can see, it found all the different ones. And look at it, creating this spreadsheet. Isn't this just wonderful? There you go, that's the first one. And we can just fast forward through all of this, right? No need to sit here for three minutes. But it goes ahead and successfully creates a spreadsheet with all the different links, pros, cons, the pricing and a personal rating that operator aka GPT4.0 came up with all in one shot. I want to add one note here, actually. Philip included a part of this prompt. In the end he told it to think and to create a plan before it went ahead with the research. It just completely ignored this. This is just GPT4.0 and it's just going to go step by step and do its thing. I think sooner than later we're going to see operator with 01 or 03 or 03 Pro or competing products from Deep Sea, I guess come out and those will be a lot more calculated in its approach. But it's incredible to see this just working with a one shot prompt. And the nice thing is you can run this prompt multiple times and it works every time. It's not like some of the competing products where it's just, just not reliable. Okay, here's another quick one. And this one is about sending messages to Slack. I wanted to highlight this one because it really shows how good Operator is at solving some of these problems. I just want to be clear, like it's not a perfect product. It's a research preview and sometimes it does get stuck in a loop. But usually with some clever prompting you can work around that and make it work and then you can save that clever prompting to the original task and have a preset that actually works. So that's incredible. In this case it's trying to use Slack and it doesn't work. As you can see, it's stuck in this loop where it keeps telling it that hey, this site is blocked and it's not going to work. It does that a few times in a row, but then it chooses another approach. I'm not sure how it found this out, but it just goes to app.slack.com then it tries the wrong approach a few more times, but then it returns to app.slack.com and realizes that hey, this works. There's a web version of this application and I can just send my message like so. And Philips Operator successfully sends a GPT4.0 written message to our team Slack, which just an amazing demo of how it just finds different ways of getting around problems. And this is just consistent amongst all the use cases that we looked at. Nice. So you can use Operator to spam your colleagues or loved ones on Autopilot. That makes me think that I might want to try out a use case where I do something like Guilfoyles AI for any Silicon Valley fans out there, he had this bot one day where he was talking to its co workers and it wasn't even him. After recording this video, I might want to see if I can replicate that with Operator. Okay, here's another interesting use case that was suggested by YouTube user William Embry 57y7 and he has asked if this could draw using any web based CAD programs. This is basically a version of a 3D software that can run in your browser. So this was another one of the examples where I approached it with a simple prompt and a more complex prompt. The simple prompt basically said access a popular web based CAD or CAD program and create a simple 3D model of a basic geometric shape. Document the process and capabilities. All right, so it went in and it basically only needed my Google login at a certain point to create a new account on some of the software softwares that it found. But I can already tell you that this task did not work at all. It actually found a bunch of these like Tinkercad and SketchUp, Web, Vectrace, Spline, Emachine Shop, which by the way I was really impressed with. It even went to some GitHub repos to find open source CAD software like this. And while on some of them it managed to log in on others it needed my help. None of these worked because all of them require hardware acceleration. In other words, they need help from your local computer because just the browser cannot run these by itself. So as you can see on the screen recording here, it always gets to the point where it opens up the software and then some error message appears that hey, you need hardware acceleration for this to work. And because this is just a simple virtual machine with very little compute backing it up and not a full fledged computer. None of these web applications worked. Nevertheless, this was a really impressive one to test, as it did find a plethora of different applications. I mean, look at that. And it figures out the interface of each one of them by itself. I think on all the ones except the GitHub it managed to get to the user interface and then it basically refused to work on this one called figura. It actually clicked the different buttons and it tries going into the canvas and drawing something up and it refuses to copyright on this one. I'm actually not sure if it's because of the hardware acceleration or it's because the software is used differently, but nevertheless it gave it a good shot. Now we're in the same task force through the anthropic prompt improver, which always does a good job of fleshing out the task into multiple steps, et cetera, as you can see right here. So it's just running the simple prompt through the prompt improver to create this and it tells you the 12 steps to the automation, including sub points and everything. So in theory this should work better. Well, that's at least what I thought. But yet again, this really elaborate prompt giving it all the details performed equally as well as the short one. Matter of fact, it found the same list of different tools worked for a total of eight minutes, whereas the other one worked for a total of 12 minutes. If you just quickly scroll over the video, it's pretty much the same thing. It opens up the different pieces of software, runs into error messages, just to eventually give up on this task. So anything that might need hardware acceleration is not going to work on this because it's just a simple virtual machine where it barely has enough hardware to run a browser. Anything like a 3D software is not going to work on that. Okay, and to round this video out, I want to give you one last example and, and this one I picked as the very last one because it just clearly demonstrates the potential here that we haven't even talked about here. The point of this video is showing you use cases that either work or break, so you get a feeling for what it can do. But I think the biggest potential here is Operator actually working with other agentic tools, whether it's other AI platforms like Claude or Midjourney in this case, or even beyond that, we're testing use cases where it's using Repl, Dot Agent to build an entire website, or where it even assembles its own automations. These apps work because they don't need hardware acceleration like the 3D modeling software. So in this case, Dom, which is responsible for the creative district in our community, ran a bunch of midjourney prompts that he included in the original prompt automatically. With Operator he basically had a bunch of variations of the same prompt with different sref codes. He just wanted to see what all the different styles would look like on it, and rather than sitting there and having to send it all manually, Operator just did it all for him. Again, this is a one shot prompt without any follow ups and it just did the damn thing. And then you could keep playing with the prompt and download these images to the local file storage and then re upload them somewhere else. Although I have to say I tried this particular workflow generating an image, saving it, and posting it somewhere else. The saving part sometimes is a bit tricky. It's really good at uploading images, but the saving step is not as good as I would like it to be yet so what's the conclusion here? Well, in my humble opinion, after a few days of really intensively packed playing with this and kicking the wheels, this thing has impressed me. As opposed to many other AI releases which are very hyped when they come out and then very often they disappoint in practice. This one was the opposite. It actually came out and the whole deep seek story just overshadowed the release of this. But in my opinion it is the most capable web navigating agent I've seen yet and I think I tried them all. Let's see, I'll run some comparisons again. Maybe I'll create a separate video if that's worth it. But the point is it's just the best product on the market right now. And I personally think that some of the use cases in this video are ones that I would actually find myself doing. Like I purposefully left out some things that I'm doing with it already and that I love. Like for example, this one that orders a basic basket of groceries from a favorite supermarket. I run this once a month to get the basics like mozzarella and bread that I like because those are things that people expect. I tried to show you some things today that may go beyond what they showed off in their blog post, but overall I would say it's been less than a week since this came out. I would not want to give this product up anymore. Is it worth $200? I guess that one is really relative because if you're making $50 an hour then I do think that this can save you four hours a month. You might have to play around a little bit, create your own presets, but that seems realistic to me. Plus, there's the upside of building skills that will transfer into the future. Other than that, it's probably overpriced for most people at this point, but I myself will stay on this journey, on this course of trying to squeeze the most out of this. And this is the first version of it where I'll be actually using this on a regular basis. Like, why would I order my groceries manually if I have my wonderful preset here? Like, I literally pressed this button yesterday and this cheese that I really like just magically appeared at my door in around 30 minutes. So if you think this could be useful to you and you want to learn more about this, then I can only recommend you check out the AI Advantage community. And we're also running weekly sessions where a team member of ours is sitting there with a copy of Operator, and if you want to run anything through it, can just leave a comment in advance or join that live event and our team will help you run some operations that might be relevant to you. All right, that's all I got for today. If we find more, I'll be following up with more videos on Operator. I honestly think this is the main thing I want to be covering on this channel. And with the Advantage moving forward, the future is here and I will see you very soon.