The Process

The premise is that workflow tools which used to take a quarter and a team can now be built solo in weeks — whether that means replacing something you've outgrown or building functionality that doesn't exist yet. Either way the test is the same: define the win up front and measure against how the work gets done today. Every engagement is built around that test, and around ending — either a tool earns its place and your team takes it over, or it doesn't and I say so.

Scoping is free. After that, each phase is priced and committed separately — you're never signing up for the program, only the next step.

How an engagement works

Phase 0 — Scoping

(~1 week, no fee). We pick two or three candidates and rank them by how clearly a win can be defined, not by the size of the prize. The first one is usually a subset of a workflow rather than a whole tool — a bounded slice that can succeed or fail fast and keeps scope honest. For each candidate, we write down the single number that defines success before anything gets built. This phase is free because I only take on candidates I'm confident will work; you shouldn't pay to find out whether the conversation surfaced one.

Phase 1 — Build

(~3–4 weeks). I build a working version of that slice and run it against how the work gets done today — an incumbent tool, a manual process, or nothing at all. What you have keeps running; nothing gets ripped out. Build velocity stops being a claim and becomes something you've watched happen.

Phase 2 — Decision

(~1 week). We measure against the number set in Phase 0. Three outcomes, all agreed in advance: it wins and we proceed to handoff; it loses and we kill it, you keep what you had, and we document why; or it's promising-but-not-yet, which buys one bounded extension and then a decision. There's no open-ended "let's keep iterating" — that's the failure mode that turns pilots into dependencies.

Phase 3 — Handoff

(wins only). We name an owner on your side, train them, and document the tool. The work ends when they're confident running it without me, not when a budget runs out. Then the loop can repeat on the next candidate, and each cycle costs less than the last because the engine persists — the later tools are mostly configuration, not construction.

  • "Nobody knows our business better than we do, but we're not a software company."

    The OpEx costs and difficulty hiring / training a support team have been completely upended by AI.

  • "We've built custom tools before. We're still paying for one nobody uses."

    Every pilot has a defined kill point. Deciding not to keep it is a valid outcome.

  • "If we build this with you, what happens when you leave?"

    Every phase ends in a handoff to a team you own.

Let’s Work Together

If you're interested in exploring a possible engagement, complete the form with a few details about your project.