method · IT professions · 7 min · interactive article ✦
Most AI projects that fail don't stumble on the technology: they stumble on framing. You start from an appealing tool in search of a problem, discover too late that the data isn't ready, that nobody thought about compliance or adoption. Here's the checklist to run before launching — tick it, measure what's missing.
take it with you · free
The checklist as a 1-page printable PDF
The 10 framing points on one A4 page, to tick in the kickoff meeting or drop into your project file. Leave your email to receive it, or download directly.
If you remember one line: never start from a technology. Start from a bounded business problem, check the data follows, frame compliance from the start, and prove the value on a short pilot before industrializing. Technology is rarely the limiting factor — framing almost always is.
Frequently asked questions
Where do you start an AI project?
With the problem, never the tool. State the business need and the measurable expected outcome before talking technology. Many AI projects fail because they start from a solution in search of a problem.
Should you build or buy an AI solution?
It depends on total cost (incl. run), lead time, vendor dependency, compliance and differentiating value. Buy for a standard need, build or adapt when AI touches your core business or sensitive data.
What compliance obligations for an AI project?
Under the EU AI Act, obligations depend on the risk level of the use. Add data protection (GDPR), accessibility and decision traceability. To frame from the start, not at the end.
Framing an AI project is a skill.
In IAPLC, you practice framing on YOUR cases — build/buy, data, governance, ROI — through 30 levels generated for your IT role.