yesil.dev
EN/TR
Back to the notes

What should the first AI sprint teach us?

Discovery should make the conditions for a useful solution visible. A working example should give the team enough evidence to make its next decision.

Engineering2 min read
An experimental prototype on a test bench with a measuring tool and successive physical samples

Choose one workflow

I suggest starting an AI project with a two-week discovery phase focused on one workflow. Where does the input originate, who handles it, where does the result go, and where does work currently wait?

Keep a measure beside the prototype

For a reporting workflow, the team could record preparation time, missing data, and review effort before trying a new solution on the same examples. If preparation becomes faster but review takes longer, that trade-off belongs in the result.

Make the next decision easier

At the end of the sprint, I would look for three things: an inspectable output, unresolved assumptions, and the conditions required to continue. Missing integration access or inadequate data should be visible in the roadmap. Discovery makes the investment decision concrete before the scope grows.