The problem
We run a profitable sports management software company with no outside investors, so every cost comes out of our own margin. With ten people, we were moving slower than we wanted and spending more than the business should.
At the same time, new AI-native companies in our space were shipping faster than we could. We weren't built that way, and we had to keep up.
The question wasn't whether to use AI. It was how much of the company's work AI could do well, what should stay with people, and where to start.
What we did
We ranked every idea before building anything
Almost anything was possible, but not everything was worth doing first. Every idea went through the same four questions: what it would return, whether the business was ready for it, how hard it would be to do well, and whether it served the goals we already had.
We went function by function, proving each one first
Sales research came first, because the return was clearest and the risk was low. Marketing followed, then engineering, then support and the back office. Each change had to work in the business, measured, before we moved on.
We held AI to the same standards as people
Every system has an owner, written standards and a review step. Nothing customer-facing goes out without a person approving it.
The principle
Start where the return is clearest, prove it, then move to the next function.
The results
The team went from ten people to three, and each of us now gets more than three times as much done.
Sales: an outbound engine researches every prospect in seconds, where it used to take 15 to 20 minutes each.
Marketing: a content engine drafts search-targeted articles, and a person approves each one before it's published.
Engineering: coding agents built a full memberships product in about a week, against a three-month estimate.
Support: an agent trained on our help centre, product documentation and past conversations answers about 70% of questions without a person.
Finance: reconciliation and bookkeeping run automatically, replacing the part-time help we used to need.
What we learned
Not every idea survived contact with real customers, and the ones that did needed an owner to keep them working. What worked a year ago is often out of date now, so we review our own setup regularly and replace what no longer earns its place.
What this means for your company
You don't need to change every function at once. Start with the one where the return is clearest, measure it, and let that result fund the next. That's what the AI Priority Roadmap does with you.
Figures from our own company. We walk through the details, privately, on a call.