Organisations are moving fast on AI experimentation and slowly on the harder question underneath it: does this use case solve a real business problem, and how will we know? Without that discipline, AI investment drifts toward whatever is fashionable rather than what moves the business forward.
Where this gets hard
- No shared definition of “value” across functions — finance wants cost reduction, product wants speed, operations wants risk reduction.
- Teams are incentivised to launch pilots, not to close them down, so nothing gets retired even when it isn't working.
- Strategy is typically set annually while AI capability shifts monthly, so plans are stale before they're approved.
- Executive sponsors change roles and initiatives are left without a clear owner.
- Boards ask “what's our AI strategy” as a status signal, which pushes leaders toward visible pilots over unglamorous, high-value groundwork.
Where to start
- Require every use case to state which business metric will move, by how much, and how it will be measured before funding is approved.
- Build a portfolio view — experiment, scale, retire — reviewed quarterly instead of a flat list of pilots reviewed annually.
- Tie AI investment cases to existing strategic pillars instead of writing a parallel “AI strategy” document.
- Give each use case one accountable owner with the authority to kill it, not just launch it.
- Separate the tactical productivity clock from the structural transformation clock so one isn't judged against the other's timeline.
A downloadable consulting document is available with a use-case prioritisation checklist, a portfolio review template, and a short self-assessment for your leadership team.
If you audited every AI initiative running in your organisation today, how many could name the business metric they're supposed to move?