I had a software development agency called Onestop Devshop. I opened it the way I open most things — as an experiment — and momentum carried me into something that became, at its peak, one of the most beautiful and easy-to-run lifestyle businesses I've ever had.
By late 2022, I had as many as 75 developers, contractors, designers, admin and ops people running through 12 time zones. BC Canada to Europe to India. Good margins. Great people. A business that basically ran itself while I surfed and hung out with my family. It was a lot of fun.
If you've been following me at all, you know what happened in 2023. Short story: my eldest son Finn died. I stopped working. Let the agency die. Clients stopped paying me and I didn't care. I got hosed by one vendor for 50,000 CHF and another for 15,000 CHF. I couldn't even bring myself to fight it. None of it mattered.
When I started to come out of the grief fog, probably mid-to-late 2024, I walked into a different world. The AI tools had arrived. Not the hype — the real ones. The ones that could actually build things.
I tried to get my dev partners excited. Use the tools. Build MVPs faster. Quote tighter. Move quicker. The faster we deliver, the more work we get. That was my pitch to them for years. They never bought it.
The Agency Incentive Problem
Here's the thing I didn't fully understand until I lived on both sides of it: the traditional software agency model is structurally pro-supplier. It's built to protect the agency, not the client.
My partners weren't dishonest. They weren't scheming to rip anyone off. But they were carrying developers. Thirty, forty people who needed to eat. Dev jobs are brutally hard to come by, especially in the age of AI when founders are experimenting with vibe coding things themselves. And when you've got a team that depends on you, your first obligation is to keep them paid.
So when a project came in that could be built in two weeks using AI tools, they'd scope it at six. Not because they were lying — because they had payroll. They had markup on every developer hour. If they quoted two weeks and delivered in two weeks, half their team sits idle. That's not a business. That's a charity, and it's a few payment cycles away from going bankrupt.
This drove me crazy. Ranting and hand-wringing couldn't get me to change the culture. I'd get on calls and beg them — guys, quote faster, lock in the client, develop as fast as possible. The work will come. I said this constantly. It was basically my mantra. They'd nod, and then the next quote would come back where I didn't feel like it should be. So we lost leads, lost deals, and I quietly just gave up.
The incentive structure made it impossible. An agency with 40 developers on bench cannot be motivated to deliver a project in two weeks. They need that project to run for three months or even better 12 months so they can bill months of developer time and pay months of salaries. Speed is a direct threat to the agency's revenue model.
This is not a moral failing. It's a structural one. The agency model was designed for a world where building software was genuinely slow and labour-intensive. It made sense when you needed 10 developers working for 6 months to build a product. But the tools have changed. The math has changed. And the model hasn't caught up. But it is 100% doomed, that's for sure.
I clearly saw the writing on the wall in early 2025 and made moves to sell the agency. That misalignment is basically why I sold the agency in 2025 to a major player in the space and walked away.
The Fractional Model Flips the Incentive
Now let me tell you how TestVentures works, because it's the opposite model.
A client comes on at a flat monthly retainer. They send me tasks through a workflow — right now I'm using Trello, eventually something more polished. A task comes in: build a website, set up an AI voice agent, automate a reporting pipeline, create a customer onboarding flow, build an internal dashboard, deploy a chatbot, audit their AI stack and tell them what's redundant. Whatever it is, it lands in the queue and I get to work.
Here's why the incentives are completely different. My motivation is to keep that client sending me tasks. If I slow-walk a website build and it takes three weeks for something that should take three days, the client thinks there's no value and moves on. They cancel. I lose the retainer.
Speed is not a threat to my model. Speed is the product. The faster I deliver, the more impressed the client is, the more tasks they send, the longer they stay. Every completed task is proof that the retainer is worth it. Every day I sit on something is a day closer to them questioning why they're paying me.
An agency wants projects to last. I want tasks to close. An agency profits from complexity. I profit from clarity and speed. An agency is motivated to keep developers busy. I'm motivated to keep clients happy.
That's the structural difference. The agency model is pro-supplier. The fractional AI ops model is pro-consumer.
Why This Works Now (And Didn't Before)
Five years ago, this model would have been impossible. One person cannot run a full-service ops function across multiple clients without the tools to back it up. You'd burn out in a month.
But the tools exist now. I can build a production website in a day using Claude as a copilot. I can set up AI voice routing in an afternoon. I can automate a client's entire reporting workflow in a few hours. I can audit a company's software stack and give them a prioritised recommendation list before lunch. The stuff that used to require a team of five working for two months can now be done by one person who knows the tools, in a fraction of the time.
That's not a boast. It's just the math. And the math is why the fractional model works.
The DesignJoy Blueprint
I didn't invent this model. A designer named Brett Williams at DesignJoy proved it works. He runs a productised design service — flat monthly fee, one request at a time, unlimited requests per month. No hourly billing. No project estimates. No scope creep negotiations. You subscribe, you send design tasks, he delivers them fast, you stay because it's better than hiring a full-time designer or dealing with an agency.
Last I checked, Brett was doing north of $100K per month. As a one-person operation. Because the model is tight: flat fee means the client's cost is predictable, unlimited requests means they feel like they're getting value, and one-at-a-time means the queue stays manageable. The constraint is what makes it work.
I'm applying the same structure to AI ops. Flat retainer. Tasks come in through a board. I deliver fast. Client stays because the value is obvious. No hourly billing. No padding. No incentive to make things take longer than they should.
When I get too busy, I stop prospecting and look to bring on a superstar to help me. I'm putting out some feelers now to prepare. When the work is slow, I start prospecting again and build a funnel where I just flip a switch and leads start to come in. If I have a handful of clients per month, two people are in good shape making good money and 'coding from the beach'. Do the clients care where we do the work or when? No, as long as we deliver, they keep sending tasks.
The Team That Actually Makes Sense
Right now it's just me doing the work. If I get slammed, I'll bring in one really good 10x operator to help me clear the queue. But I don't need 75 developers to mark up on. I need two, maybe three people who are exceptional at driving AI tools, building agents, and delivering work that makes clients want to send more.
The dream team for this model is small and specific. A visionary who handles client relationships, sales, and marketing — that's me. An AI agent specialist who can run six projects simultaneously and ship fast. A developer who can crank through the technical queue when the agent work isn't enough. Maybe a designer or generalist to round it out.
Four people. All getting paid well. All running what amounts to a lifestyle business with real revenue and real impact. You fire the clients who drain you, you lean into the ones who energise you, and you repeat. Expand when it makes sense. Keep it tight when it doesn't.
That's the model. Not 75 people across 12 time zones hoping the next project estimate doesn't get cut. Four people who are so good at what they do that clients never want to leave.
The Numbers
Revenue: $0. Clients: 1. Prospects: 4.
Still building. But the model is clear now in a way it wasn't when I was running the agency. I spent years trying to make a pro-supplier model work for the client. It doesn't. The incentives fight you the whole way.
This time the incentives are on the same side as the client. That's the difference. That's everything.
Day 14 of 365.