Running My Life and My Business on AI Agents

00:07MC: All right, we’ve got one last presenter tonight. This one’s going to be Mr. Fleming Slone, the founder of Autonomous Agent AI Consulting. Please give it up one time for Mr. Fleming Slone.

00:29Fleming: Thank you. So, this is normally what I do—I’m not going to be as good as the last speaker, but before I get into it, I just want to thank everybody for coming tonight. This is our first one of these meetups, which I’ll talk about in a bit, but we’re definitely trying to have these every month. We’ve got them scheduled for the rest of the year, so I’ll keep coming back out and hopefully this will continue to grow.

Also, a special thanks to Curtis for hosting tonight, and thanks again to Tim for speaking. And some of you may have seen him floating around with the camera—we’ve got a guy in the back who’s with Atlanta News First here in town. If y’all follow Atlanta News First… (aside) I didn’t use AI for that one, but it might be for the next one. So yeah—thank y’all for making this happen.

01:47 I was going to talk a little bit about how I ended up here, and how I used AI to get here. I’m also going to be showing the Claude Desktop app. So if you’re not as technical, and you want to get in on some of this stuff but you’re thinking “how have I used Claude Code? I don’t use the terminal”—this is the way, right here. With the Claude Desktop app you can access pretty much everything we’re talking about tonight. It even works on your phone—if you don’t have a laptop handy, you can run Claude Code and everything else right from your phone.

02:37 So, on April 15th—what’s that, about 45 days ago?—I got this email that said, “Hey, remember that AI group you went to a year and a half ago? That guy disappeared, we don’t know what happened. You want to take over an AI group?” And I’d just bought this house earlier this year, so I’ve got all this space—“come here, let’s do AI together.” And I’m thinking, how am I going to do this with a full-time job and everything else? And of course the answer is AI.

03:17 I know a lot of people here are familiar with Claude Code—show of hands, how many of you are on the plan? … Okay, about half the room. So if you’re not on it, I’d recommend it. What I’m showing here is Cowork—this is a feature from Anthropic, built into the software, that can go off and do tasks you ask it to while you’re working on something else.

03:55 You can see just in my recent history—I’ve had it going out and finding me volunteer opportunities, finding a wedding gift for a friend, things like that. You don’t need to do all of this stuff, but I’m working a full-time job, so I just send it out. What it actually does is automate the browser—just like a user, it goes in, clicks around, and finds things in the menus. I never go digging through menus myself; that stuff drives me crazy.

04:37 So here you can see we’ve scheduled an entire year of these events. Once I took over the group, I just had that Cowork feature in Claude go out and make all of it—it basically filled a full schedule for the year, put the details in, and it’s been sending out messages with updates. I checked some of it, but Cowork did basically all of it.

05:11 And yes—the logo and pretty much everything you see here was AI-generated. I think people talk about the negative aspects, and they’re real, but I think it’s important to recognize the positive ones too—because without this, this group basically wouldn’t exist. I wouldn’t have had the time to do all this. So that’s one thing I look to: it’s creating community by letting people accomplish more.

05:56 At the same time, consider how easy it is to create events and websites like this—a bad actor could do the exact same thing. The same stuff we’re using for good could obviously be used for bad.

06:16 But as part of this—we’ve got this new group now, and we need a website. I’m a front-end developer, 20 years; that’s what I do, building websites like this. Before, how long do you think this would take? At least a couple of weeks for a design and an active build. This was basically done in an afternoon. Now there’s a full website with an integrated schedule, running a worker in the background—a fully professional site in one afternoon.

07:01 Then I got a little carried away and made a website for the house too. So now the house has a name and its own site. You’ve got to have fun with it. And then—“well, we need a Patreon, right?”—so I had the AI go make us a Patreon. You can become a member for only $50 a month. (Laughs) That’s a joke—don’t sue us for copyright.

07:45 But this is just one example. I basically run my life with this now. I went and got a haircut the other day—for the haircut I just pulled up Cowork. Some people fully automate this stuff, but if you don’t want full automation—say you still want to go in and look at the appointments yourself—you can just pull up Cowork (or Claude Code, if you’re comfortable with that). For someone less technical, Cowork is great: “Hey, I need these groceries from Sam’s Club,” and it places my Sam’s Club order and I just go pick it up.

08:33 The implications are pretty interesting, because it’s improving my life and letting me accomplish so much more—but at the same time, we’re basically living the Jevons paradox, right? So, interesting things to think about.

08:53 Here are more examples from just the last month. Here’s one of those dashboards you build and definitely should never push to production… I’ve never done that. But you can see that across this network of sites—through Autonomous Agent AI, my consulting side business—we’re running 21 sites. Most of it is automated: it sends us downtime alerts, but it also goes out and diagnoses problems on its own. If the system determines a site is down, it can autonomously go out and fix it.

09:55 And I’ll end on this—this one’s more of the technical piece from the last month, for me as a developer. This is a phone-booking agent. You can call up my business and it will talk to you and handle everything. It uses ElevenLabs and things like that. But what’s really impressive is that this system was built and QA’d fully autonomously. I never wrote a line of this code. I never even looked at a line of this code. I just kept improving the agents surrounding the code until the system was complete.

10:48 This project was roughly a hundred thousand lines of code—which, being optimistic, would take a person years. It was done in a day and a half, for less than $150. And, you know… it’s a little scary, because I make more than $150. But at the same time, I think it’s important to recognize that these things are possible, and to invest our time at the higher level. Instead of being in the weeds writing code by hand—we still enjoy reading it—we have to leave some of that behind. To me, this is the future of programming: building systems around your desired outcome that ensure it.

12:06 If you’re familiar with GitHub, you can see—I forget the exact count, 60-something PRs—it’s opening and reviewing its own PRs.

12:22 And now, a demo. Let me call Autonomous Agent AI on speakerphone. [Live demo—Fleming calls the booking line; the agent answers, takes his request, and books an appointment.]

13:00 That’s the whole call. This system—zero human-written code, zero human QA—is now booking appointments on its own. Those are some of the things I wanted to show off from the last month. I don’t have too much more—any questions?

13:19 Oh—I guess I didn’t do my bio. I’ve been in tech for 19 years now, full-stack developer, which is relevant here: the first time I was laid off by AI was 2019, when machine learning took my job at a computer-vision / camera-security company. So it’s not the first time—which kind of makes me feel a little better, since I’ve been through it before. But yeah, that’s pretty much it for me. Any questions?

Q&A

13:55Q: As a developer being creative and experimenting—building your own stuff on the side—how does that affect your appetite for entrepreneurship, versus the corporate game?

Fleming: I’d say it’s greatly increased my appetite for entrepreneurship. I’ve always been the “idea guy,” and now I can actually execute on the stuff—so I’m just in here going at it. It’s definitely increased my entrepreneurial drive.

14:46Q: Can you share a product you’ve built—a real one?

Fleming: My nine-to-five doesn’t use AI as much; they’re not an AI-forward company, so we use it behind the scenes. That’s another reason for the side hustle—I find this stuff infinitely fascinating. Most of our work is around integrations. Our primary product is this network of sites, which we monetize different ways—some with advertising, some by selling backlinks, and so on. We’re not selling an “AI product” per se; we’re running a portfolio of publishing sites, and AI is how they’re run. Increasingly, clients see it and go, “oh, this could benefit my business”—so the integrations side is catching on.

16:43Q: Did you live on the BeltLine before you moved here?

Fleming: I did not live on the BeltLine. This is very new for me. I’ve seen pretty much everything you can imagine out of that window in about four months—you wouldn’t think it’d be that crazy.

17:06Q: What’s the thing that gets you most excited every time you use it?

Fleming: The most exciting thing to me is that phone system, because it was the first time I achieved full autonomy—I’m pretty much out of the loop on it now. As for other things—this was a very early project we built with AI: a hurricane evacuation-zone planner.

18:08 What it does is aggregate data from state-level agencies—hurricane/emergency management agencies—and present it in a way people can search. This was done with tools like Claude and design tooling. It’s got an interactive chat for emergency preparedness, things like that—a pretty cool project.

19:16Q: Do you use any off-the-shelf tools for this?

Fleming: We don’t use much like that. Pretty much everything we use, we’re building ourselves as our needs dictate—because now it’s possible. We’ve got a few things we pay for, but otherwise we build exactly what we need—even our own CRM.

19:51 And that’s where I think Salesforce might be in trouble. If people can build a product that’s more tailored to their needs, without spending $50 million to create it like Salesforce did—companies starting now can build these products and become major disruptors. You can build things cheaper and more tailored than what’s out there.

20:33Q: What’s your goal for this whole meetup series? This is the first one—congratulations.

Fleming: Thank you. First off—Tim is a big inspiration for this series; I don’t think he even knows that. Tim is one of the best sources of AI information around, and I thought, there have to be other people out there doing all kinds of interesting things. I’d wanted to do this for a while: it’s a way to share what people have been up to, because things are moving so fast it’s hard to keep track.

21:19 The second goal—we don’t know how this all ends. Regardless, it’s good to build community, to be around each other. Strength in numbers. Maybe we won’t solve everything in one day, but I just want to get people together and see what we want to build. It’s been fun.

22:04Q: What about compliance and confidentiality with your customers’ data, given how automated the work is?

Fleming: Our clients so far are very small businesses, so there isn’t much in the way of formal compliance. But on protecting information in that pipeline—data is going out, certainly. Right now we’re sending it to Opus, so for the moment, like the previous speaker said, I wouldn’t send truly sensitive information to it. Part of the answer, in my mind, might eventually be running your own models—your own servers, keeping it in-house—as the way to provide better privacy. But right now we’re not dealing with anything dangerous enough to worry about it too much: standard encryption and so on for the PII we store.

23:35Q: Do you play around with open-source models yourself?

Fleming: A little. I still need the hardware to really run them well—it’s hard to get a good GPU. I’ve played with them; the last time I was really digging in was probably around DeepSeek. There again, you don’t necessarily want to route customer data through it. So yeah—it’s definitely time to revisit that.

24:36 That is pretty much it, guys. I did want to thank you all for coming out to the first of hopefully many. I learned a lot, and I hope you enjoyed learning some things too. Thank you all again for coming out—you’re more than welcome to hang out. Enjoy the rest of your evening, and I’ll see you all next time. Thank you.