AgenticWorks

I connected with a developer, early in 2025, who was managing a small team of junior developers. We saw the same opening: businesses were desperate to “do something with AI” and had no idea where to start.

The Thesis

His idea — sell agentic-workflow and automation builds to small businesses first, use the client work to find the patterns worth productizing, then build the software version once we actually knew what it needed to do. I’d seen this sequencing work before, more than once, and I still think it’s right: a service business forces you to solve someone else’s real problem, on their timeline, before you ever guess at a roadmap.

What I Built

We had the pieces in place — a partner, a working dev team, and a pitch aimed at small businesses that wanted AI workflows and automation without hiring for it. I’d worked closely with dev agencies for years before this — hired them, evaluated proposals, watched founders run them — so I understood the model’s real pressure points: hiring, margins, scope creep, the gap between selling a contract and delivering one.

What Happened

The strategy wasn’t the problem. The market underneath it had already started moving. AI made it faster and cheaper than ever for a small business to get a workable agent or automation built — by an agency, or increasingly by someone with little coding background just prompting their way to something good enough. The scarce skill was shifting from “can someone build this” to “can someone get it in front of the right buyer and keep them there.”

My partner and I had different instincts about where to lean into that shift. He leaned toward organic growth — build something good and let word of mouth carry it. My instinct is to treat growth as a system: a hypothesis, a test, a measurement, a repeat. Neither instinct is wrong; they’re just different ways of thinking about the same problem, and a two-person agency doesn’t have room to run both approaches at once. We each ended up pulling toward the version of this business we were built to run.

What I Learned

Service-first before software is still the right sequencing — but watch the market underneath it, not just your own execution. As building gets commoditized, the constraint moves downstream, and a service business built on “we can build it for you” has to keep re-checking whether that’s still the scarce part.

• Pressure-test how a partner thinks about growth before you build a pitch together, not after. A shared theory of how demand actually gets built matters as much as shared technical skill.
• A service business is a bet on how you want to spend your time, not just on the market. That’s worth checking as early as the strategy itself.
• When building stops being the bottleneck, distribution and judgment become the whole game — plan for that shift instead of being surprised by it.