Start with the approach, not the technology.
The problem
An AI project that starts from "we need to do something with AI" risks stalling: a proof of concept that ends up in a drawer, or a pilot that never reaches production. So start from "we have a specific problem that AI can solve."
What actually works
Three things help:
1. A concrete problem Not "we want to use AI" but "our order desk retypes every order by hand, and that goes wrong sometimes." The more specific the problem, the easier it is to measure whether the solution works.
2. Measurable outcomes Define what success means upfront. Not "work smarter" but "fewer orders that sit for a day."
3. An iterative approach Start small, measure, improve. No big-bang implementations. The order intake system I built has been running in production since July 2026: every order is pre-checked and released by a staff member.
The takeaway
AI is a tool, not a strategy. Start with the problem, not the technology.