Why The Old Company Model Breaks

SPEAKER_01 0:00

Welcome back to Inspire AI, the podcast where we explore how artificial intelligence is changing the way we work, lead, build, and think about what's possible. There's an assumption we've carried for generations about building a company. You have an idea. You find customers, and as the business grows, you hire people. Marketing, sales, operations, engineering, customer support, you name it. But what if that isn't the default anymore? My guest today, Nicole Patton, has spent the last several months testing a very different model. She built a successful business in roughly six months with essentially one employee or stuff. By using Claude and creating AI agents to take on work that traditionally would have required a team. And Nicole isn't approaching this simply as someone who discovered AI yesterday. She's an engineer and a product builder who spent seven years at Google and today works with organizations on Cloud implementation, AI systems, and AI fluency. So in this conversation, we're gonna dig in and try to understand the architecture of a one-person AI company. What work did she delegate? What did she refuse to delegate? What broke? And what actually worked? And perhaps most importantly, if one person with a collection of intelligent agents can operate something that once required an entire team, what does that mean for how we build companies in the future?

Nicole’s Path From Google To Solo

SPEAKER_01 1:20

Let's dig in. Nicole Patton, welcome to Inspire AI.

SPEAKER_00 1:23

Thank you so much, Jason. Glad to be here.

SPEAKER_01 1:26

Awesome. So why don't you start by telling the audience what brings you here today?

SPEAKER_00 1:30

Yeah, so I met you, Jason, at AI Ready RBA, the annual conference, which was great. And since then, things have just taken off as far as my AI business. So I'm an AI consultant focusing on Claude, and I'd love to tell you about how I run the whole thing with myself and Claude.

SPEAKER_01 1:48

Excellent. Yeah, I understand it's been quite a journey for you over, you know, six months plus essentially you're you know one person building your business. You were inspired by the new technology gains of efficiency. And why don't you take us back to the beginning when you thought about this and you realized that you didn't need to build a traditional team, a company to do what you wanted to do?

SPEAKER_00 2:12

Yeah, absolutely. So I left Google and it was just I've just thrown in with the wolves, basically, like you just had to figure it out. Then it was at a time, obviously, when AI businesses were really blowing up. And so originally I started and I wanted to, you know, they tell you to niche down and focus on one technology. So I started with AI voice agents, actually. That I quickly realized that was not going anywhere. And then people started coming to me asking about consulting, just talking, I was very vocal about my use of claws. So that's pretty much what started the whole the successful side of the business. And yeah, I just started to get really, really passionate about automating all the things. So starting with like the most basic thing, it really the the first automation was a stripe payment leading to an invoice creation in QuickBooks. And so I didn't have to do all of that, and then just slowly or maybe quickly thinking about every little piece of my business that I didn't like to do, and then automating that until I am here where I have basically like my whole team is pretty much agents with me in the

Agents As Interns Across The Business

SPEAKER_00 3:29

loop.

SPEAKER_01 3:29

Oh, yeah. Very good. So tell us what does that look like? What what jobs are you personally doing? And what jobs do you effectively have assigned out to your agents?

SPEAKER_00 3:40

Yeah, I mean, I guess it's pretty once you start automating, you start realizing exactly what needs a human in the loop. So obviously it's me and an agent or agents for pretty much everything. So I think of them more as interns. So I have a a sales agent that goes out and scouts the web, looks for leads and just different opportunities for me, sends them back to me. Obviously, they're not all great. So me as the human has to decide what I should reach out to and what I should not. But same with like with marketing, with with finance, I have, you know, a finance end of month like PL. So obviously that's very easy for a robot to do. I just review it. But yeah, so pretty much, I guess where it really comes down to a human being necessary is really in client success. I do a lot of training, a lot of consulting, and that one-on-one with the client like really requires a human touch to be able to understand how to teach and how to make it so these concepts land. And that I think that's why my business is successful because that is my focus. While everything else, you know, all the noise in a business is automated. So yeah, that's basically where the the human touch really is required.

SPEAKER_01 5:00

Very

Lead Scouting Workflows And Iteration

SPEAKER_01 5:01

cool. So your first example there of you know, going out into the wild and looking up what kind of leads exist for your business. Can you walk us through that workflow a little bit more? I'd like to know how did you evolve that and you know what what opportunities do you think exist for further enhancements to AI to allow that to be a little bit more successful for you?

SPEAKER_00 5:26

Yeah, definitely. So everything has been an iterative process. There was never a time that I created an agent and it was perfect on a first run. So it's been a lot of me looking at the output and saying, here's what's wrong with it, here's what's right with it. Yes, these leads would be are great. These other opportunities, obviously, you know, AI is finding things that are in the past or are just completely not the right fit. But the more that I tell it what's right and what's wrong, the better it gets. And I think that's the the biggest thing, I think, with with AI now. And I I think probably even down the road, no matter how how good AI gets, unless until it really understands you, it's not going to understand what is good and what is bad. So you you have to teach it that, especially if you're you're building it yourself. Obviously, if you buy a you know a third-party app or something that's already been trained on this, maybe it will will be a little bit better and more on point. But yeah, so it took a lot of iteration and me telling it, this is bad, this is good, until we're at the point where I'm like, wow, you're you're you're doing great.

SPEAKER_01 6:33

Awesome. Okay. Yeah. So you're you're teaching your intern what you like, what you don't, and what what what success looks like. Uh do you use do you use loops? Do you use like workflow agent, agentic loops? Like, what does that look like architecturally speaking for that specific use case?

SPEAKER_00 6:53

Yeah. So

Monthly Website Optimization Agent Loop

SPEAKER_00 6:54

for this use case, I do. So obviously, you know, every time it goes out to the the web, it comes back with the results. We either I, myself, and another agent will review it and say, you know, this is this is good, this is bad. I think the the the best example of a loop I have right now is a an agent that actually searches the web for like a competitive analysis. So it does that basically monthly. It also does that with my my ICP. So looking at my basically ICP avatar, it looks at the metrics of my website. So Okay, what's ICP?

SPEAKER_01 7:32

Sorry.

SPEAKER_00 7:33

Oh, sorry. My ideal client profile. So it looks at who I'm targeting, right? So who do I want to buy for me? So it's doing all these. I have agents, a different agent searching the web and compiling all this information. And then basically I have an orchestration agent that looks at all of that and it's like, here's what your competitors are doing. Here's what your ideal client is looking for, here's what your website says. Here's how we suggest that you change your website, change your your offer, you know, and I have that running once a month. So it's like this whole loop that is basically, and then it will just create a PR and be like, review this. And I'm like, all right, that's good. And then I'm like, oh, wait, this sounds too much like AI over like whatever. But so it's basically like doing all this research for me and then coming back and then proposing these changes to my website, my offer. And then I'm like, yeah, looks good.

SPEAKER_01 8:28

Oh wow, that's that's amazing. That's exactly what I was looking for there. So there's there's a huge difference between like Chat GPT and Claude actually designing a business around AI. When did you cross the line of using AI operationally speaking versus managing and running your entire business?

SPEAKER_00 8:49

Yeah, I think just, you know, coming out of the gates with just really like no backup and kind of just like I don't have the the means to start hiring. And then the more that you start using AI, you know, I started to be like, you know, just small things, incrementally being like, okay, there's, you know, we can do the stripe, the finance thing. And then from there, just fully automating. So I don't know if there was actually really a line, or if like straight from off the bat, I was just like, all right, it's me and Claude, you know, we're gonna get the same design. We have nobody else.

SPEAKER_01 9:23

Right on.

SPEAKER_00 9:23

So yeah, yeah. I didn't really have any other options.

SPEAKER_01 9:27

Right, right, right. There's a a phrase there I'm trying to pull from my memory about necessity is the mother of invention, right?

SPEAKER_00 9:36

Yes, that's that's a good one. Yeah, that it is that pushed you. That's very true. Yeah, yeah.

SPEAKER_01 9:41

Awesome.

Automation Failures And Client-Facing Risk

SPEAKER_01 9:42

So, what did you initially think an agent could handle that turned out to be a terrible idea? Give us an example of a big failure for yourself.

SPEAKER_00 9:50

I guess so, just right off the top of my head when I first started, I set up this automation that when somebody booked with me or sent like a calendar request, it like triggered this whole flow of basically like setting up the engagement. And it got that wrong quite a bit, and it was very embarrassing. So I had like existing clients reach out to me, like, yeah, by the way, I got this email that's you know, is like obviously for like leads or or for like a whole different engagement. So yeah, I took a step back with that one, and I want to be like every anytime anything a client sees, I am extremely careful of. So yeah, I think that was a big lesson because that was very embarrassing.

SPEAKER_01 10:37

That's fantastic. I I was just thinking about how I need to automate my outreach because the popularity of this podcast is growing, and we get one or two requests a week right now through through our website at AI ReadyRVA to be you know volunteering on a podcast and then random like cold calls. And I need to automate all that because small business-wise, I'm I'm a one-person show for this podcast. Right. So yeah, I'll I'll take a lesson from you there.

SPEAKER_00 11:12

Yeah, we can talk.

SPEAKER_01 11:14

How do you decide what gets delegated as an agent versus what Nicole needs to own?

SPEAKER_00 11:19

Well, anything that has a clear process, so going from trigger to this is exactly what happens no matter what, to end result that goes to an agent. Anything like customer success, client success, negotiating with a client, figuring out dates, that all needs a human in the loop, deciding on the curriculum, all of that. So yeah, I found out very quickly what I should delegate and what really needs me. And yeah, you saw that in the example with the agent that went rogue and sent something to an existing client about signing up.

SPEAKER_01 11:55

Absolutely. So,

The Economics Of An AI-Run Company

SPEAKER_01 11:57

what are the economics of a company like yours? Like when software can give one person capabilities that previously required several employees, like what fundamentally changes? Can you talk a little bit about that?

SPEAKER_00 12:09

Yeah, I mean, the economics of it are really the beauty. So instead of hiring a full team, actually last night I asked Claude exactly for you know this workflow, how many people would I need to hire? And what can I delegate to an agent? And it really gave me a really good overview of if I didn't have AI, how many people I'd have to hire to get the job done. And instead, I'm spending about seven to eight hundred dollars a month on tokens. So yeah, you can imagine the margins there are pretty high.

SPEAKER_01 12:45

Tremendous. Yeah, I won't ask you what Claude said. I'm sure it elaborated a little bit, but nonetheless, it's probably pretty close to what you're doing with your business.

SPEAKER_00 12:55

So yeah, yeah. It basically I asked it to break down exactly who I would need to hire, you know, pre-AI. And then now that it's post-AI, like let's see what an agent can do and what a human needs to do. And it was basically all agents. So yeah.

SPEAKER_01 13:11

Yeah.

SPEAKER_00 13:12

Exciting.

SPEAKER_01 13:12

All right. So you know what mature engineering looks like because of your time at Google? What if you were to start a company today, and would you find there are roles that are deliberately not something you would hire for? Oh, yeah. Because you can build them with an agent first?

SPEAKER_00 13:29

Yeah, 100%. I mean, I think one of the the first things that I would that most people I think probably hire for is a VA. And that is completely replaced with agents in my case. So I don't need somebody following up to the client, reminding me when I need to reach out or or do XYZ. So the VA had basically fully been replaced, in my opinion, in my business specifically. And I I'd say that I can do the job now of several engineers. I would probably have a person doing more QA and strategy. But as far as the if you give the agents a goal and tell them what good looks like, you will have that done, you know, by the end of the day. Whereas it would have taken a team of engineers, myself included, probably a week or so.

SPEAKER_01 14:26

So yeah. So you you've been referring to your agents as interns. And when you think about engineers and the talent and the experience that they come with, generally speaking, you can't run a company of with engineers of just interns. So tell us a little bit more about your feelings of like or what what's your perspective on on the maturity level of these engineers? How much of the role can they uh autonomously operate within? And what what is it that you absolutely as an engineering thought leader need to take action on still?

Orchestration Skills And Writing Clear PRDs

SPEAKER_00 15:06

Yeah, absolutely. So I am at an advantage, obviously, because I have the vocabulary. I know how to, you know, if I had a team of interns, I'd know how to speak to them to appropriately delegate and describe the task and their goal. So that is a real advantage because explaining appropriately the task, the goal to the agents, that really takes you like leaps ahead. So I think that the human in the loop myself is that delegation layer, that orchestration layer, and knowing how to be very deliberate with what you you tell the agents to do. So, you know, always creating like a PRD. Like an agent can't do something if you can't, if you don't even know what it is, you know. So, like really defining like this is the goal, this is what I want it to look like, but still giving it the agency to make its own decisions because sometimes it's actually, you know, things of things that you haven't. So it really becomes more human-computer collaboration at that point, you know, you're giving it ideas, it's coming back with ideas, you're saying the the overall goal. And I think that the more that you work with it, then the more that that relationship builds and the more that it understands what you want and what good looks like. So it takes some time, but yeah.

SPEAKER_01 16:29

Yeah. I so I know you mentioned your orchestration. I think you said orchestration, or at least that's the way I related to it.

SPEAKER_00 16:36

Yeah, yeah, yeah.

SPEAKER_01 16:37

Is able to, you know, assess and tell another agent to do something. Do you have other you know intern agents talking to each other and making decisions amongst themselves and sharing the feedback? And is that is that something that you've matured your your business to do?

SPEAKER_00 16:54

Yeah, definitely. And I'm I'm slowly growing that. So I think the most mature one that I mentioned was this kind of like website optimization agent. So basically there's one agent who's the orchestration. You can, and I would give that a higher model, maybe Fable or Opus 5.5. And then the other agents would be it's kind of like a hub and spoke model. So the other agents would be all right, go and do the competitor analysis, look across the web, see who is in the same industry doing the same thing. The other agent would be look at the ICP, you know, look at the ideal clients that would be coming to Elevate Online. Like, who are they? What do they want? What are they looking for? I have another agent that's looking at all the signals from, you know, Google Search Console, Microsoft Clarity, looking at like what my website is saying. And so basically all of those agents will go out and do their own thing. Let's say it's on the first of every month, and they will come back to the orchestration agent in this hub and spoke model who will look at all the data that's come in and then be, you know, that's the orchestration layer that will say, all right, this is the executive decision. This is what we need to change on our website to optimize it to make these sales for based on our comp, you know, what our competitors are doing, what our our clients are looking for, and what our messaging is. So that's basically a a simple breakdown of kind of what that model looks like.

SPEAKER_01 18:17

I see. Okay. Yeah, I was interviewing somebody the other day, and imagine you have a salesperson and a marketing person, and they each have an agent. So instead of your hub and spoke model with your orchestrator, your fable model, you got two people and they own agents and they do agentic things for those people. And then their agents talk to each other. And so the model there that I'm drawing a comparison to to see if you had the example is instead of uh two people, you have two interns. And are they communicating with each other directly yet?

Why Hub And Spoke Beats Agent Chatter

SPEAKER_00 18:55

Yeah. So that's that's interesting. And in all of the systems that I've seen, so you know, I've I've taken the anthropic training and they always recommend this more of this hub and spoke model. So they are very anti agents directly talking to each other, and agent has one specific job, and that job is to gather information and report it back to whoever's the decision maker. And in that model, the the decision maker, their one job is to make the decision. So I yeah, I think that agents talking to each other, I think that they need to have one job. So I think whenever they have two jobs, that it's like go out and gather the information and then make a decision with another agent, that's too complicated for for the AI at this point, I believe.

SPEAKER_01 19:46

Fantastic. I I appreciate your your opinion there. I it's it's not every day that I get to to hear multiple perspectives on that subject. So I figured I would I would push on that just a little bit. All of this is very product provocative, of course.

Three-Year Outlook And Responsible AI

SPEAKER_01 20:01

So let's just advance three years from now. You've got a one-person company. What is it capable of that it's not capable of today?

SPEAKER_00 20:11

Yeah, that's that's interesting. And I mean, I'd love to know where things are three years from today, because I mean, it's incredible how quickly things have evolved. I mean, you can, you know, a year ago things were much different than they are today. So I would like to think that three years from today, that these interns that I'm talking about are are so evolved that I can just trust them with handling tasks and really actually maintaining a whole role, maybe even to the point of hiring, you know. So I'm just thinking very, you know, three years is a long time in in the in the world of AI.

SPEAKER_01 20:50

So yeah, really gets it absolutely. So you you in three years from now they're no longer interns. Maybe they've been promoted.

SPEAKER_00 20:59

Yeah, yeah. Maybe they've been hire like running the sales department, right? Maybe, maybe, maybe I have a agent that's running the sales department that's hiring, you know, figuring out the other interns or agents, you know. Um now now you've got me like really excited thinking about the three years from now with AI.

SPEAKER_01 21:18

So I I feel like I'm a bit of a realist, but I I imagine like a futurist. So yeah, you know, I I like to think about how the you know turning the corner here on the the advancements with the the capabilities is gonna improve our life. Yeah. And so I you know, I like to steer in that direction when I think about it. And absolutely so that you know it's not all doom and gloom. You know, there's there's no skynet here unless unless that's something that you're really aiming for for whatever reason. Right. And and it sounds like you're being very responsible, very human in the loop, managed, orchestrator, agent, run, and making sure that all your interactions with your customers are are clean and intentional.

unknown 22:05

Yeah.

SPEAKER_01 22:06

So I'm sure they really appreciate that. And it sounds like a great advice for anyone out there looking to start their own business.

SPEAKER_00 22:13

Yeah, I think that now people are so familiar with AI and it's very obvious when something's AI written, or you know, the claudisms are very obvious. So it becomes very clear to your clients very quickly when you're over-relying on AI.

SPEAKER_01 22:31

Indeed, it does. Well, thank you so much, Nicole. It's been a really awesome conversation. I've learned a lot from you today, and I've been inspired by your your stories here of how you built your business and where you where you've taken it so far, and what we hope for from the future as well.

SPEAKER_00 22:52

Yeah, great. Thank you so much, Jason. It's been a really exciting time to get into AI, start a business app with AI, and also you know, teach AI to everybody so that they can do the same thing. So yeah, thank you so much for for having me. It's been a great conversation.

SPEAKER_01 23:09

Yeah. Let's draw to a close here. Tell tell everybody on the on the podcast where they can find you if they're looking for your particular expertise.

SPEAKER_00 23:19

Yeah, absolutely. So easy to remember claudtraining.com. I'm on Instagram as the Claude Queen. Yeah, that's that's where you can find me. I'm teaching Claude to the masses. Yeah.

SPEAKER_01 23:31

So you would prefer Instagram?

SPEAKER_00 23:33

Yeah, LinkedIn, Instagram, yeah, or just email me directly. Nicole that elevateonline.com. Yeah.

SPEAKER_01 23:39

ElevateOnline. Awesome. Thank you so much for sharing it today. Yeah. We'll see you around AI ReadyR VI.

SPEAKER_00 23:47

Absolutely. Yeah, thank you so much, Jason.

SPEAKER_01 23:49

All right, cheers.

SPEAKER_00 23:50

All right, bye.