Everyday AI Podcast – An AI and ChatGPT Podcast - EP 573: ChatGPT Agent Mode Overview: Real use cases and 3 worthwhile tips

Episode Date: July 23, 2025

ChatGPT Agent Mode is here. If you're wondering what types of use-cases we're using internally and some tips to get you ahead of the curve....we gotchyu. Make sure to join us as we put AI to Work this Wednesday. Try Gemini 2.5 Flash! Sign up at  AIStudio.google.com to get started. Newsletter: Sign up for our free daily newsletterMore on this Episode: Episode PageJoin the discussion: Thoughts on this? Join the convo and connect with other AI leaders on LinkedIn.Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineupWebsite: YourEverydayAI.comEmail The Show: [e-mail] with Jordan on LinkedInTopics Covered in This Episode:ChatGPT Agent Mode Overview & RolloutDifferences: AI Agents vs. Agentic WorkflowsChatGPT Agent Mode Hands-On DemoVirtual Desktop & Browser Capabilities ExplainedChatGPT Agent Use Cases for Business AutomationReal-World Agent Mode Example: Podcast AnalyticsPros and Cons of ChatGPT Agent ModeThree Key Tips for ChatGPT Agent SuccessAgent Mode Security, Privacy, and RisksCompetitive Landscape: OpenAI, Microsoft, Google AgentsTimestamps:00:00 "Exploring ChatGPT's Agent Mode"03:42 "ChatGPT Agent Mode Overview"07:14 "Enhanced ChatGPT Capabilities"13:31 "True AI Agents Unveiled"17:14 Spotify Podcast Metrics Challenge19:58 Podcast Retention Rate Analysis24:17 "New Tech Feature Faces Bugs"26:49 Google's Project Mariner Innovation29:43 Meeting Prep Automation Tool34:26 ChatGPT Agent Mode Benefits35:21 Real-Time Chatbot Interaction Tips41:51 ChatGPT Strengths and Weaknesses45:10 AI Agents for Truck Drivers?Keywords:ChatGPT Agent Mode, AI agent, OpenAI, generative AI, agentic model, virtual environment, agent-powered workflow, pro users, Plus plan, subscription rollout, Microsoft 365 Copilot, Google Gemini, Google Agent Space, enterprise AI, computer using agent, virtual desktop, API connection, terminal access, file upload, CSV analysis, Buzzsprout, Spotify podcast analytics, Apple Podcasts, podcast retention, data aggregation, live demo, automation, hands-on AI, multi-platform data extraction, PowerPoint creation, spreadsheet automation, connectors integration, privacy and data security, browser control, prompt engineering, context window, deep research mode, AI-powered spreadsheet, meeting prep automation, CRM data enrichment, repetitive task automation, manual data entry, multi-step workflow, virtual browser, business intelligence automation, AI-driven presentation, user activity log, iterative prompt refinement, SaaS integration, troubleshooting AI agents, future of AI agentsSend Everyday AI and Jordan a text message. (We can't reply back unless you leave contact info)

This is the Everyday AI Show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life. Meet Firefly AI Assistant, now live in Adobe Firefly, the all-in-one creative AI studio. Just describe what you want to create and the assistant handles the rest, orchestrating multi-step workflows across Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome, the assistant accelerates execution. You think you've probably used an AI agent before. But there's a good chance you haven't.

You were probably using an agentic model or an AI powered workflow. But now that's probably going to change because we've seen big AI tech now usher in real actual agents. agents that are capable of completing tasks in a virtual environment, like they have a virtual computer like an intern, and they can carry out complex tasks, even if you don't give them the best directions, even if you aren't quite sure how the tasks can be accomplished. So today, we're going to be going over ChatGPT's new agent mode

and giving it a little bit of a deeper dive in our special weekly segment called Putting AI, to work on Wednesdays, where once a week we go over real, actual, and live use cases of AI technology. I show you how I'm using a certain piece of generative AI and encourage you and give you different ways that you can use it too. All right, I'm excited for today's episode as we go over ChatGPT's new agent mode, and I hope you are too. But if you're new here, welcome and what's going on? My name is Jordan Wilson and welcome to Everyday AI.

This is your daily live stream podcast and free daily newsletter, helping everyday business leaders like you and me, not just learn what's happening in the world of AI, but how we can leverage all this information to grow our companies and our careers. It starts here with the realest thing in AI, unscripted, unedited, live streaming podcast, showing you all how it's done. But if you want to be the smartest person in AI in your department or at your company, the cheat code is ready? Write this down. your everyday AI.com. That's our website. So we're going to be in our free daily newsletter that you need to go sign up for. We're going to be recapping the highlights of today's episode,

as well as keeping you up to date with everything else that you need to know in the world of AI. So yeah, if you do want that AI news, make sure to go check the newsletter. So chat GPT's agent mode has been out now for about five days to pro users. All right. So if you're on that $200 a month plan, you've had access now, like myself, for a couple of days. And Chad GPT just, or Open AI, just tweeted out a couple of hours ago, actually,

that they've started rolling out this new agent mode to chat GPT plus subscribers. And they said the rollout will take a couple of days. So even if you are a $20 a month subscriber, if you look right now, you might not have access. I don't have access in any of my $20 a month plus accounts. but I know the majority people are on that plan. So that's kind of the biggest question we've been getting it. It's like, hey, I don't see this. Is it broken?

Is it not coming out? Open AI did say they're starting the rollout to those $20 a month plus users, FYI. So in today's show, we're going to give you a hands-on overview of chat GPT's new agent in their agent mode. We're going to talk about some real life use cases that I think you'll find valuable. And at the end, we're going to reveal three tips that will make chat. ChhpT agent perform much better each and every time. All right. And we did do a brief recap of Chad Chupt's agent mode, a couple of hours after it came out

on episode 570. So we're not going to be going over all the basics. So if you want to know all the basics, make sure you go listen to last week's episode. That's episode 570. All right. So before we jump in and we're going to do hands-on demos, what could go wrong, I first wanted to zoom out and talk about the broader picture. So Open AI obviously isn't the only player in this game, but it does look like they're the biggest player to go public first with

their agent, or at least they're kind of main agent. So Microsoft already has similar technology. They have computer using agents, but it's not quite the same thing. But Microsoft, some Microsoft executives have hinted that they will eventually support Chad GPT's agent mode. So with the Microsoft OpenAI partnership. Most of OpenAI's biggest innovations eventually find their way into the Microsoft 365 co-pilot ecosystem. So keep that mind. Google, we've been seeing reports now for a couple of months that Google will be rolling out

their agent mode inside Geminaut. So that's a little different. They have their whole agentic platform called AgentSpace, which hasn't rolled out to a ton of businesses yet unless you are an enterprise company. but they will be rolling out their broader agent mode inside Gemini as well. And there's plenty. And when I say plenty, there's literally hundreds of other kind of AI agents that do something very similar to what we're going to be going over today in chat. ChbD's agent mode, such as Manis, Jen Spark, too many to count.

I will say this. And I actually had a keynote presentation last night at the Microsoft office here in downtown Chicago. I said this. If you're an enterprise company, you probably shouldn't be. be using an AI startup. Yeah, they're shiny. Yeah, they look nice. They might do some things better than chat GPT's agent mode or when Google's version comes

out or when Microsoft's version comes out. But if you are an enterprise company, I would probably not be putting too much of your long term or even your midterm kind of AI strategy behind an AI startup that could go poof overnight. You know, nothing against these startups. That's just happened over the last three years. since I've been covering AI technology. So always be cognizant of that.

So a real quick hands-on. And for our podcast audience listening, check out your show notes. People always ask, you know, oh, is there a video version of this? Yes, there is. It's on our website, your everyday AI.com.

All right. So I'm going to be walking you through and we're going to be doing a live demo here in a second. But here's the basics. All right. The good thing is this works all inside of ChadGBT. This is essentially combining ChadGPT's operator. which is able to handle more complex websites that require interaction, which is important.

All right. And then it also incorporates their deep research mode, which is more of a text crawl, where it can go over dozens of different websites at once, as well as computer using capabilities, the ability to create PowerPoints, spreadsheets, all of that. We already did that in an earlier episode, so go listen to the last one. So if you want to use chat chTPT, the good thing is it, or sorry, if you want to use agent mode, the good thing is it's inside chat chat. GPD, which is great because previously, you know, as an example, operator, open AI's first computer

using agent wasn't inside of chat GBT. It was actually on its own subdomain. So you couldn't work with everything else that Chad GPD had to offer. So the good thing now is you can't. So if you are on that plus account, right, I advise you log out, log back in, you know, check every day or so. And then you'll see under the tools, a button, you'll see a new agent mode pop up. And it will also tell you how many queries that you have left.

So on the $200 a month pro plan, you have 400 queries. On the plus plan, you have 40. All right. And then you'll just select agent mode. And then from there, and I'll show you this live. It's going to go ahead and kick up a virtual environment. It can run actions in the terminal connect to public APIs. It can do literally almost anything.

I like to say, this is an intern, right? You can almost do anything an intern could. And then you can obviously, there's some elements of, traceability, observability. So you can go back and kind of see, watch a little video recording after it's done of exactly what the agent did. All right. What could go wrong? What could go wrong?

Live stream audience, what's up? Let's take a look live. Good morning. Brian. Brian, beat me to the live stream. Brian, thank you for showing up. Michelle, good morning. Michael, joining from YouTube.

Brea, joining from Chicago. Like myself. Marie, good to see you. Everyone else. Rolando Jose, if you have any questions, get them in now, Cecilia, holding the down for Chicago as well. All right, we're going to look live at agent mode. If you have questions, let me know also live stream audience. If anything goes wrong, right, which stuff does a lot.

And I don't want to miss it. Just please let me know. All right. So I'm going to be sharing my screen now. All right. So if you can see live stream audience, if you could let me know, I would appreciate it. All right, so I've already started something here, and I'll tell you exactly what I did and why I did it, but I want to get this prompt running first. I don't think it's going to finish in time, but we're going to give it a shot.

All right. And actually, I need to upload a file here. All right. You also give me a quick second. I'm going to upload a file. There we go. And I'm going to get agent mode going. All right.

So let me describe as this hopefully works and kicks off here. All right. A lot of people are excited for this one. All right. Murray said we can see. Thanks, Marie. All right. So here's what's going on.

Here's what I already did. Actually, let me actually just say the prompt that I put in first. And then I'll explain what I already did in this instance. So I said, I'm giving it some background. Simple stuff, nothing crazy. And again, this is my example. Put an AI to work on Wednesday. Think of your own example. So I said, I run the everyday AI podcast. I'm focused on growth. One thing I'm looking at is better understanding and improving my retention. The podcast is hosted on Buzz Sprout and distributed to dozens of other platforms. The biggest are Spotify and Apple Podcasts. I've uploaded stats from my BuzzSpout account. So that's the file that I just uploaded. In your web, browser. You're logged into my Buzzsprout account as well as my Spotify account and Apple podcast account. So that's what I just did in the kind of context window up here. I started to get the agent going. Then I clicked take over. Okay. So you can pause and you can take over. So why did I take over for the agent? Well, it has its own virtual computer and I needed to go log into my Spotify account

that has all the podcast stats. And I needed to log into my Spotify account. And I needed to log, into the Apple podcast account that has all this podcast stats as well as our Buzz Sprout account. So before I started, I took it over. So I did this initially once or twice on Thursday and Friday. So I do like any good cooking show would. I do have a version of the cake that's already been baked because I also don't want this episode to accidentally go, you know, 45 minutes because I know people like Jeremiah are at the gym right now.

And I don't want to keep you on, you know, the elliptical. longer than you have to. So I do have this one done already. So we'll give it some time to see if it will actually go through. And I want to show you guys under the hood exactly how agent mode works. But one of the biggest tips already, right, is go in and log into these services. Yes, you always need to keep, you know, privacy, data security, all of those things in mind. So, you know, if, I don't know, if you're dealing with PHI, PII, right, probably not something

that you should go log into, talk to your leadership team first, right? So if there is a super secure system that you log into via the web and you're like, oh, I'm going to go ahead and give chat ChpD's agent this. Well, can it do it? Yes. Should you? I don't know. Go, go talk to your people, right? So I'm just getting that precursor out there. So what I'm having chat Chb TvT log into, it's just a bunch of numbers, right? There's no PII, you know, personal information, health information, you know, credit card information, anything like that. And Open AI has said and warned everyone, which I appreciate that they did in their live stream. Agents can be dangerous.

They can go off the rails, right? This is actually their first model or mode that Chad GPT or sorry, OpenAI essentially slapped a high risk label on. Keep that in mind. Again, I'm going to draw this comparison again. This is brand new. These capabilities, again, unless you've been using, you know, one of the better, uh, true agent startups. We haven't had this, right? A lot of people are, you know, on, on LinkedIn and Twitter,

they're sharing, you know, custom GPTs or, you know, Zapier automations or N8N workflows. And they're saying, look at these agents. Those aren't agents. Those are AI powered workflows. An agent is literally like an intern who can literally do anything in a good way, but can also do anything in a bad way as well. You know, it has terminal access. It can connect to third party APIs. It has a browser. It can go do anything that it wants, even things that are outside of your instructions, which again, usually is good because people give very vague directions. And sometimes they're asking for a lot. So the agent will have to go and, you know, go down some paths that you didn't think it would travel and neither did it. Right. But this is brand new. This isn't like deep research. This isn't like,

you know, chat, GPTs operatives. mode. This is something else completely. This is literally, I like to say an intern because it's not that great yet. It is a little slow, right? But this is an intern that works 24-7, 365. It doesn't stop, even though it is slow, even though it is a little clunky right now. When it works, it is very capable and it can create Excel spreadsheets and it can create PowerPoint presentations as well. So in this prompt, that's all I really did, right? Then I'm asking it specifically what I want. So I told it where to find the retention stats on Apple and Spotify. And then I told it, I've made some slight changes to the podcast recently. I want to look at my retention rate,

essentially from the first couple of weeks of June of last year versus the first couple of weeks of, or sorry, July of last year versus the first couple of weeks of July from this year. So what this would entail is a ton of manual work. Because one thing I don't like about Spotify, and that's our main platform. We're usually like a top 10, top 15 tech podcasts on Spotify. We have, I don't know, like 100,000 or almost 100,000 subscribers on Spotify. That's our big platform. Spotify's stats stink, especially when it comes to retention.

So what retention is and what I'm, you know, having operator work for, and let's see, hopefully it's going here. All right, good. Let's see. All right. Unfortunately, let me see. Could you share a spreadsheet? I'm going to say, please try again.

All right. We ran into a little bug. And I'm going to say if Spotify doesn't have that data, look in Apple Podcasts. All right. Again, generative AI, even agent mode is generative. I ran this exact same prompt on Friday. It worked very well. This time, it didn't work.

Right. And that's a pro and a con of using any gender to VI tool, especially agents. So it already, I scrolled down, it ran into a little roadblock. That's okay. But essentially what I'm looking for is I've, I've tried to make some changes to the podcast over the last couple of months. I'm trying to make it a little more concise, even though some episodes like this are a little hard, right? Because I want you guys to really see under the hood.

I'm trying to be a little tighter, right? With delivery, not ramble on as much, which is hard for me, if you listen to the podcast a lot. Right. I'm trying to do a better job also in the beginning to say, hey, here's what we're going to go over. And then at the end saying, hey, as a recap, here's what we went over. So I'm trying to see if that improves retention.

And unfortunately, in Spotify, there's no way to do that just with one metric. I have to individually go into every single podcast, three or four clicks, and then manually grab those stats. I can't export it. And that's a big thing for me right now, running my business. So as you think of your own use cases, and I'm going to give you good use case examples, think of that thing that right now, whether it's a spreadsheet, a PowerPoint presentation, navigating around the web in some platform that you use that's cumbersome.

And you're like, man, all I'm doing all day is I'm just clicking a lot. I'm reading a lot of information. And, you know, I'm only using my brain for 10% of the time as an example. Think of those tasks, those mundane tasks. This for me is a very mundane task. We have 560 plus episodes and to do three or four clicks in two or three different platforms over and over and over to try to say, you know, oh, when I do this type of show, are more people listening all the way through, right?

If 20,000 people start an episode and only 2,000 people end it, that's bad, right? And then I would have to go in and look at that episode. I don't remember, oh, this episode had a high retention rate. Let me go back and see what it was. I'm going to have to go back and watch it, look at the transcript, etc. This is a great use case for an agent. It can go. It can juggle the data.

It can go and browse a website, unless in this example, when I'm trying to do it, and it's doing a terrible job. All right, let's see. I'm going to say, please navigate and find it yourself. How did I know this would happen? How did I know this would happen? It's doing a poor job. Let me try again here, y'all.

let's try from scratch. Let's try from scratch. And then everyone can see exactly how agent mode works. Well, hopefully. All right. So I don't know why we ran into a little, a little hiccup there. So again, I'm going to chatgbt.com. I'm clicking the tools.

I'm clicking agent mode. I'm pacing that prompt in. Hopefully at least we can get this one working correctly. All right. So now you'll see it says setting up my desktop. Actually, I lied. I did that wrong again. Here I am burning through all my agent mode queries just for you because I forgot to upload

my CSV file with all the Buzz Sprout information. There we go. We're going to give this one one more try. But like I said, worst case is we're going to be able to jump in and look at the results because I've already done this once. But anyways, let me finish exactly what I'm asking for. So I'm asking for retention rates and then buy Corti. Hopefully that's how you say it, right?

But different podcast platforms, they say, oh, you know, in the first part of your episode, you know, you obviously start with 100% retention rate. And then, you know, the second kind of quartile, it goes to, you know, 80%. And then it goes to 60% is to go through this over hundreds of episodes and then to go back and research why, you know, a certain episode is higher or lower. Maybe I rambled on. Maybe I had a great guess. Maybe I had a not so great guess. Maybe it was a hot trending topic. I don't know, but an agent can go and find those out.

And again, I'm telling it to look at the retention rate from the first couple of weeks of July 24 to compare it to July 2025. Then I'm saying record all the stats and put them in a spreadsheet. Then I'm saying put together a short visual presentation. Please take your time. Yada, yada, yada. All right. So let's see.

All right. Now at least our live stream audience can see. Chad GPT agent mode is actually getting to work. You can see it here working live. So now that is actually working, let's take a very quick 30 second break for a word from our sponsors. This podcast is supported by Google. Hi folks, Paige Bailey here from the Google DeepMind Devrel team. For our developers out there, we know there's a constant tradeoff between model intelligence, speed, and cost.

Gemini 2.5 Flash aims right at that challenge. It's got the speed you expect from Flash, but with upgraded reasoning power. And crucially, we've added controls like setting thinking budgets. so you can decide how much reasoning to apply, optimizing for latency and costs. So try out Gemini 2.5 flash at AIS Studio.gov.com and let us know what you build. All right. So we are back and now looking at agent mode go. So this is great. In FYI, I am now watching this agent work in real time. And if you did use operator before, you'll notice this is much faster than operator, right?

in kind of the big bump here in why this is extremely exciting and a true agent. A lot of previous computer using agents, the way that they operated is they took essentially screenshots each and every time. And you can see how that would take a long time. This is completely different. This isn't taking a screenshot and then navigating. This has an actual virtual environment, right? That is logged into and stays logged into the services.

that I log into. All right. It looks like it's having some errors, some Spotify errors, obviously, fun times. All right. But you'll see it's actually working. It's buggy right now, right? And it's probably going to continue to be buggy for a couple of weeks. But luckily, I did run this once or twice, this exact prompt right when it came out.

I do know that chat GPT plus users on that $20 a month plan are getting access to this. So I do assume that it's going to get hit pretty hard. Luis here asking, Good morning from Brazil. How to enable the agent mode in a plus subscription, unable to see the option here. Different countries. We've had this in our newsletter.

And I'll have to check in on this. I do believe different countries have different rollouts, especially if you're in the EU, things might be a little bit delayed. But like I said, the plus for those $20 a month subscribers, it is slowly rolling out over the next couple of days. I would assume it's going to take at least a week for it to work.

But what that does mean for everyone, if you're going to rush and try it out now, it's going to be super buggy, right? Because now it was great from, you know, Thursday when they first announced it, you know, through the weekend. I was doing it on my phone when I'm in the car. I'm using Asian mode all the time. But yeah, now that, you know, everyone is getting access, right? Tens of millions of people of plus subscribers are used.

it's going to get a little buggy. All right. We're going to let this continue to go on. And for our podcast audience, what's happening now is I'm seeing the agent right now is in my Spotify account. It's going through episode by episode. And it's also kind of narrating, right? So it says, I am currently waiting for the episode list to load once it finishes, right? And sometimes you can't even read the whole thing. But essentially, as it like clicks on things, as a new screen loads, it kind of says what it's doing. doing what it's thinking, right? And then at any time on this virtual environment, I can go, it's kind of small,

but in the upper right hand corner, there's three dots. You know, all the big tech companies do this. I get it. You want to have a nice UI, U.S. But we need to make this a little more prominent because most people don't know. All right. So I can click that and then I can choose activity mode. And then activity mode, which is great.

It actually has a screenshot. So, you know, as I'm watching the desktop view, I can't always read everything quick enough because it actually moves very fast, right, compared to operator. It's not taking a screenshot, even though it's showing me a screenshot after. But the way it's navigating the web, it has a virtual desktop. It has a virtual browser. It has a virtual terminal. So it can move very quickly, at least compared to operator.

So I can go click the activity tab if I want to go back in time and see and verify everything, which you need to be doing with agents. right? I talk about the 1% misalignment issue all the time and how that can be compounding when you're working, especially in a multi-agent system. So you want to be able to go back and trace this and observe it. So you can do that in the, that's kind of the action or sorry, the activity mode. And then you can toggle between the activity mode and then the desktop mode. You can click stop at any time and you can also take over the browser. And the reason why you might want to take over the browser is then to go log into your accounts, right?

In this case, I logged into my Buzzsprout, Spotify, and Apple. So it can go and go do a lot of my work for me, right? I do hope that all the other companies take a page out of Google's book. One thing that I love, not in their Google agent mode, but in Google's Project Mariner, is it has this like teach and repeat thing, right? So if you have these mundane tasks, it might be hard to kind of prompt engineer your way to it, especially if there's kind of some intricacies and details. So I love that option in Google's Project Mariner.

And I hope it rolls out to other platforms as well, where you can essentially record an action that you do, multi-step action. And then Project Mariner can then run it on repeat. So hopefully, you know, Chad Chivete and Microsoft and the other big players will have an option like that in the future. because I think what agent mode is good for, it's a lot of those very repetitive, mundane, technical tasks, such as researching, logging into multiple accounts, grabbing data, putting together that data in a presentation, right,

and in a spreadsheet. And that's exactly what I'm doing here in this live demo. Live stream audience, if you do have any questions, get them in now. I'm going to go over a couple other things. And then we're going to wrap the show up. So if you have questions, let's talk about it. But let's talk about some real use case examples. The first is a weekly executive dashboard.

So this can pull your team's metrics from multiple sources, analyze trends, create formatted slide deck with key insights and recommendations. That can do four, eight, ten hours of manual work in minutes. That's another big one. This is something, right, back when I had, you know, my marketing agency actually up and running, right? We actually still have one client on that side, on the Accelerant agency side. But this is, this was one of our employees jobs was to log in, right? We'd have to log into Google Analytics, Google Search Council, you know,

meta, Google ads, you know, hot jar, right? Like five, 10, 15 different services that we were doing a lot of in-depth marketing work for these companies. I had one person, we had one person that this was their primary job, logging into all of these different systems because there wasn't a unified way to bring all this together. They would pull all of this data, put it in a spreadsheet, and then make a presentation out of it. This is literally one of the best real use cases for chat chbtee agent mode. Number two, meeting prep automation. So you can give access to your calendar, your Gmail via the connectors portion as well. So it can access, and that's another thing, right, I uploaded a file. That's one

example, you can still use all of the capabilities of chat GPT, including its connectors, its integration, right? So you don't even have to do every single thing by uploading files or navigating to a web. You can use the connectors and integrators as well. But one is meeting prep automation. This can read your calendar. It can research attendees, attendees companies for recent news, poll relevant project files, and create talking points brief. This is another thing that I do a lot. right a company reaches out and they're like hey jordan we want you to train 5,000 of our employees on chat gpte right so literally the first thing i do is you know who's ever on that email i'll usually manually research them because i want to know a little bit about what they do in their role i'll

research the company i'll research recent news right so then when i sit down uh you know i'll research some of their competitors i'll research trends in their industry so when i sit down with someone and they're like hey you know because it's not cheap and they're like you know, we want you to do a huge training for us. So I'm not unprepared. I've had to, you know, up until recently, I've had to do all this work manually. That's another great task that you can just have this new chat, GPT, ancient mode do for you.

Another one, vendor research and RFP response. The other thing, which I didn't really talk about, this has a virtual computer, which means it can not only click, it can fill out forms, right? Yes, there is still, you got to have. the expertise driving the loop. And there is some risk to giving an agent the ability to submit information that is important on forms. But maybe it's not super important. Or maybe there's RFPs that you just don't, you and your team just don't have time to submit to. And you're like, man, this is manual work. Right. There has to be a better way. Well, here you go. Uh, you know,

maybe you should be taking the most important RFPs, but maybe you want to give, uh, ChachPD's agent mode a try. So you can do, you can find qualified vendors, compare pricing and features across different websites. Excuse me. Y'all, sorry. I'm still sick. I've had a weird cough for a week. So yeah, sorry if I'm a little horsey and still nastily coughing into the mic every once in a while. But you can go compare pricing and features across 10 plus websites, fill out initial inquiry forms, and create comparison spreadsheets with recommendations. One more use case, CRM data enrichment automation. Everyone could be doing something like this, right? So this is something if you have a lead

list, all right. Hey, here's warm leads or maybe your company, you know, maybe you're doing outbound sales, your business development rep, your sales development rep, and you have a lead list or an account list. Then you have to go manually do a bunch of research and update a CRM. All right. Well, I don't know. Give that to chat. You VD agent mode. Again, make sure you go through the proper data protocols, you know, privacy, security, all that, right? So, but if your powers that be, say it's okay, I would do it, right? Observe it. Don't have it, you know, work on your most important accounts or do your entire job at first. Give it a small task like I just did, right? First, but this, you can take your lead list, research each company across LinkedIn websites,

log that data into a CRM. Yes, it can type, it can save information and then populate missing fields with current data. Again, this can turn weeks of manual data entry into hours of works with a real agent. All right. Here we go. And if you do have any more questions, y'all, please get them in now. I'll try to, if you do have any questions, answer them at the end. It can just be my opinion on something.

What should I use it for? If you have questions, get a minute. But I now want to go over three tips. to make chat GPTs agent mode perform much better each time. Tip number one, hands on the wheel. You saw an example of this. I showed you. Take over the browser control.

A lot of people aren't doing this because they don't know, right? Because it's that little, you know, user interface, user experience thing. You know, if you watch Open AI's live stream, you obviously know you can take it over, but a lot of people don't know. Click that takeover button. It's a browser, right? It's based, I believe, on chromium. So you can open new tabs.

You have to enable copy and paste. But once you do that, you can enable copy and paste. And for the most part, I've logged into probably 10, 15 different websites. It seems to follow the exact same login state as it would on your local machine, which is really cool. Right. So it's not like if you shut down your browser and you go back the next day, you're going to have to re-log into everything.

Right. There's certain pieces of. that I use that just kind of stay logged in for, I don't know, a couple days a week. Then there are certain pieces of software I use that you have to re-log in every time. What I've seen inside chat, you see the agent mode in the login state, it kind of follows what it does. So it's almost like you have a dedicated Chrome environment that saves your login state, which again, keep privacy in mind.

All right. But then the other thing, when you're, when I'm talking about hands on the is you can just start chatting with the agent, which is much different, right? Because if you're in previously, if you were in operator mode or deep research, you can't really interrupt. The great thing about the agent, just like an intern, you can stop over and you can say, hey, actually, I told you the wrong date. And I know you're, you know, five minutes, 10 minutes in your research,

but I want you to just pick up right where you left off and then go back in, if you made any mistakes on this date that I gave you, update that accordingly. that's great. It doesn't stop the process. It will, you know, maybe you just are adding more details and you're like, oh, I actually, hey, in addition to retention rate, I want you to give me a country by country breakdown, right? Or maybe you did give it wrong information as you're observing and then you wanted to go back and just update, you know, one of the five different fields that you told it. So that's the other thing.

You can just start chatting, which, uh, let me just go ahead and show our, our live stream audience this because it, like you actually. wouldn't know, right? Because if you're looking at my screen, and if you use Chad Shoebt a lot, in the, uh, in the input box, it just has this stop sign, right? And so you're probably like, oh, there's not an arrow. So I can't talk to the agent. So once I click and start typing and I say, uh, like actually, you know, update the countries as well. I'm not going to do this, right? But then I can click the send button. So a lot of people don't know. that you can actually go in mid in the middle of your agents work and update, modify, give new directions, et cetera.

And you need to look like I showed you. You need to look at the activity log and the desktop so you can better understand. You really need to have hands of the wheel. Tip number two, use the context, right? Use the context window of chat. Youvety. Upload your files. Use the data in integrations, right?

User connectors like Canva, Gmail, Google Calendar, Google Drive, Outlook, email. etc. This is the big, the big plus side that it is inside of ChadGBT and not on its own, you know, subdomain like operator, you know, dot ChadGBT.com or Sora.com. It is inside chat Chb-T. So you really need to use the context window of chat chpt to its fullest advantage. Upload your files, you can use connectors, et cetera. Don't start typing, right? First, have a conversation with the agent, although the agent is very slow, right? Like you're using a specialized version, I believe of 03. So it's slow, right?

So if you just say, like, hey, I have this task for you. What do you think? Ask me questions. It's not going to get back to you in, you know, five seconds. It might take two, three, four, five minutes. However, over the long run, especially if you're going to be doing a repeating task, I actually think it's worth it first to have a conversation. And then tip three, you need to fine, tune repetitive.

agent runs. Okay. So again, that's what I advise you to do. Don't just do something that you would do once, right? What is that one manual repetitive process that you do every single day, that you do every single week, that there's really no other way to automate with AI, right? And this is where this combination of deep research operator, having a virtual desktop, a virtual browser, command line tool, the ability to create files, this is where it comes in really handy because it can do a lot of different tasks and keep that context across multiple domains, multiple programs, and create something of business value. So you should be breaking out your most repetitive manual tasks, either daily or weekly.

The very first time you do it, don't stop at midway, right? Just observe, all right? The second run, change nothing and take notes. And then on your third run, that's when you should start iterating, right? Oh, halfway through interrupt it. So that's what I, especially if you haven't done this before, first run, just sit back and watch it. All right, even if you're like, oh, I did something wrong. I could have made it better.

Second run, just take notes. Manual. Yeah, I know that sounds scary. And then your third run, start iterating on the prompts or the inputs. All right. So that's a wrap, but let's go in like any cook show, cooking show should and see if it actually finish. So you'll see here, fortunately, unfortunately, this one's still running, right? Which means after some of those earlier hiccups, it's actually working now. So, uh, I, I had a feeling we,

you know, demo demons, uh, got me again. Um, and obviously this is rolling out to now tens of millions of new users. Uh, so keep that in mind. Don't try this once this week and be like, ah, it stinks. I'm never using it again. It's going to get crush. Open AIs, GPUs are going to melt. Make sure check back next week the week after. Demand will cool down. I'm sure opening I will throw more resources behind this. So this version didn't finish. No worries.

All right. So now I'm showing a finished version. And you'll see here it has the podcast retention stats. Not CSV. And I can download that and open it as well. All right. There we go. So I know you can't see this because I'm only sharing the browser.

but I will, let's see here. I will just take a quick screenshot so you all can see it did the work and pretty, pretty impressive. All right. So here is our retention screenshot. I literally just downloaded the file, screenshot it, uploaded it to the cloud. So here we go. Median consumption, the ratio of the median versus the duration, the first quartile, second

quartile, third quartile, completion percentage, right? It did a great job. right. And then I can go back in and it also created a nice PowerPoint presentation that I can download and modify as well, you know, including it has some graphs, some explanations, right, some tables, some comparisons. You know, again, it's not the most beautiful, you know, chart in the world, but it's pretty, or sorry, PowerPoint presentation, but it did the job, right? I think a lot of the other AI agents out there create better looking PowerPoints, but I'm guessing that chat GPT or sorry, open AI will continue to improve this.

All right. So couple quick takeaways, some pros and cons, right? It can be slow, right? If you've used, especially if you've used other agents or if you've never used an agent before and you're expecting it to go as fast as you, you might say this is slow. You might say this is fast, right? So that's a con. It's not as fast as some of the other AI agents out there.

From a presentation standpoint, it doesn't create the most beautiful PowerPoints, right? Whereas other AI agents actually create really good looking PowerPoints. So there's some cons, but some pros. I mean, chat chadipt agent mode does great handling complex research, data analysis, and across multiple domains, right? things that finding information that's hard to find, right, logging into your systems, right? It can apply filters, select things from drop downs, right? It can do complex research, data entry, and file creation tasks just like an intern would. So if you have that mentality, this isn't going to replace your work, all right?

This is the first version, right? Like the way I talk about it is remember like the first. version of chat chbtee it wasn't good it wasn't good the first version of uh google bard the first version of microsoft coat right the first version of anything when you look back at it not that good uh right so keep that in mind this is a brand new actual employee right whereas before when we're talking about you know AI and even as these models become more capable more agentic right more talking about um automating manual work we're usually talking about chunks right Hey, it can't do a whole project front to back,

but if you break that project into five, six, seven different chunks, then it can take different pieces of this project. This is the first time that we've had from the big players, a publicly available agent that can do the whole project. You have to be a little patient. You have to learn, right? And you have to understand how this thing works. But overall, it's impressive.

All right. Let's see here. A couple questions. Let's go ahead and knock some out. All right. Let's see. Someone asking, is Agent Out for Plus users? It started rolling out yesterday.

I'm guessing it'll be done by the end of the week. Again, check your country to see if you have availability. Another question here from the YouTube machine. For PowerPoint presentations, isn't Gamma or Presentee, the software to use? Yeah. I mean, Gamma makes better. presentations visually, but as an example, what I just had it do, it can't do that. Gamma, presenti, right, can't go log into all of my websites.

It can't go do the research, right? So if you have a big, if you already have text ready to go into a presentation, if you're only looking at creating a presentation, right, as that one bite-sized task, I don't think that agent mode is the best for that one thing. But it's great at being able to go find the information. research across multiple domains do things that are computationally challenging and put it all together for you in one shot. So if you're only looking for a presentation, if you're just looking for something to

create a PowerPoint for you, there's better tools out there. And then another question, how would AI agents work for truck drivers? Well, I'm not quite sure. But think of whatever manual work the truck driver may have to do. in front of a computer. That's the other thing. You know, when you think of what can agent mode or future agent modes from Microsoft, Google, et cetera, what can it do for you?

Think what do you spend so much time on sitting in front of a computer and manually doing? What are those repetitive tasks? All right. So I hope today's show was helpful. We gave a more in-depth hands-on overview of chat Chb-chpti-ChpT's agent mode, gave you some real use cases, both my own going over at, live and four different example use cases, as well as giving you three worthwhile tips on how to get the most out of chat GPT agent mode. All right. So I hope this was helpful. If you want to see

more on chat GPT agent mode, let me know. Again, we did go over another episode going over the basics. So if you just listen to this one and it was a little too advanced or if you have more questions, make sure to go back and listen to that episode from last week. Was this helpful? If so, If you're listening on the podcast, please subscribe to the show. Leave us a rating. I'd really appreciate that. If you'd, if you want to share this, I'd even like that more, right? A lot of people are like, hey, Jordan, I heard from someone last night when I was doing this keynote for Chicago Tech Week that said, hey, I've been sharing your, you know, sharing your podcast with all my coworkers, family and friends.

I appreciate, man, my heart was very warm in that instance. So please, if this was helpful, we put a lot of time. to demystifying all of these AI advancements in tech, even when I'm a little sick, right? Even when I don't feel good, I know that you all out there are looking for a reliable source of unbiased information on how to use these tech tools.

So if you could share this with your network, I'd appreciate it. Also go to your everyday AI.com. Sign up for the free daily newsletter. Thanks for tuning in. Hope to see you back tomorrow. And every day for more everyday AI. Thanks y'all.

Meet Firefly AI assistant. Now live in Adobe Firefly. the Allman One Creative AI Studio. Just describe what you want to create in your own words and the assistant handles the rest, orchestrating multi-step workflows across Adobe Creative Cloud apps, including Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome while the assistant accelerates execution. Stand control with the ability to step in and refine at any time.

See it today at firefly.adobie.com. And that's a wrap for today's edition of Everyday AI. Hi, thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going. For a little more AI magic, visit Your EverydayAI.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.