The Question That Comes Too Late
Most AI conversations in organisations begin at the same point: the tool. A board mandate arrives. An investor asks the question. A competitor is mentioned. And within weeks, the organisation is evaluating platforms, sitting through vendor demos, and shortlisting options. The question of which tool to use gets answered before a more important question has been asked: what decision, in this business, are we actually trying to improve? That sequencing problem is not a minor inefficiency. It determines whether the investment produces an outcome or produces activity.
What the Tool Cannot Tell You
AI tools do not arrive with a map of your organisation's decision landscape. They do not know which of your processes is the actual bottleneck, which of your data sources is reliable enough to act on, or which outcomes your leadership team will accept as evidence of success. Those answers have to come from inside the organisation before any tool enters the picture. When they do not, the tool gets implemented against an undefined problem. It may produce outputs. Those outputs may look impressive in a presentation. But if no one defined what 'working' looks like, there is no way to know whether it has worked. Pilots close. Vendors move on. The organisation is left with a licence it does not know how to justify.
The Right Order of Operations
The organisations making genuine progress on AI — not just announcing it — share a common discipline. They started with the business problem, not the technology. They asked: where are we making decisions badly, and what would it take to make them better? From that question, the data requirements become clear. From the data requirements, the tool selection becomes straightforward. The implementation has a defined target. The outcome has a measurable benchmark. This is not a slower path to AI adoption. It is a more direct one. The work done before the tool is selected is the work that makes everything after it defensible.
