Self-test · 60 seconds
Is it an AI problem?
Four questions. They take a minute here and an hour with us, and they sort real AI problems from the ones better solved another way. Answer honestly; the verdicts appear as you go.
Can you name the decision or outcome that actually gets better if this succeeds?
Good. A nameable outcome is the foundation every successful project shares.
Stop here first. Name what is actually wrong: the decision that is too slow, the delay a customer feels, the task wasting skilled people. Technology chosen before the problem is named serves the vendor, not you.
Would a simpler tool do the same job?
Then it may genuinely need AI: variance, language or judgement at a scale rules cannot handle.
Boring technology wins. An integration, a report, a workflow rule: cheap, reliable, already understood by your teams. This is a happy result, not a failure.
Is the underlying process sound?
Good. Automation amplifies whatever it touches, so it should touch something worth amplifying.
Fix the process first. Put AI on top of a broken process and you get the same mess, faster. This is a structural problem: people, ownership, process.
Who owns this after the project team leaves?
That name is the single best predictor of the capability still working in a year.
Unowned capability decays into shelfware. Decide the owner before the technology, not after.
A genuine AI problem survives all four. Most briefs do not, and finding out in a minute beats finding out at the end of a six-month engagement.