Prioritizing AI use cases without the hype
The right use case is the one that fits the operating model and creates leverage the business can own — not the one that makes the best demo. A field-tested filter for the shortlist.
A practical filter
Good prioritization starts with the economic shape of the work, not the novelty of the technology.
The best candidates are repetitive, visible and tied to a decision or handoff the business can improve. If we cannot describe the before/after in one sentence that a CFO can read, the use case is not ready for a pilot — it is ready for a discovery conversation.
The economic shape test
Before anything else: does the workflow cost money, take time or create risk in a measurable way today? If the answer is 'kind of', the economics are not sharp enough to defend a pilot budget. If the answer is 'yes, we can point at the line item', the use case earns a place on the shortlist.
The economic shape also tells us what kind of result the pilot should produce. Cost-shaped problems want a unit cost delta. Time-shaped problems want a cycle-time delta. Risk-shaped problems want a control point with an audit trail. The result format has to match the problem shape — otherwise the week-twelve conversation falls apart.
Ranking candidates without a scoring matrix
We avoid scoring matrices. They look rigorous and produce false confidence — every parameter is subjective, the weights are guesses, and the output is a ranked list that feels mathematical but isn't.
Instead we run three filter questions. One: can the team sketch the workflow in five minutes with the operators in the room? Two: is there a KPI already being tracked that will move if the pilot works? Three: can we write the acceptance criterion in one sentence? Use cases that clear all three go to the shortlist. Everything else goes to the parking lot.
What to do with the ones that don't make the cut
Most AI shortlists have two or three strong candidates and eight weak ones. The weak ones are not trash — they are signal. They tell us where the operating model is ambiguous, where KPIs are missing, or where the economics haven't been framed yet.
We document the parking-lot items with the reason each one was deferred and the condition that would unlock it. That document becomes the roadmap for the second engagement — once the first pilot has proven the commercial shape of the partnership.
More from the field notes
The AI pilot blueprint
The shape of an AI pilot decides its fate before anyone writes a prompt. Here is the engagement we run — phase by phase — and why every slot is engineered to close on a handover, not to keep the advisor in the room.
Read moreWhy AI projects fail at the process layer
A useful AI initiative still collapses if the underlying process is unclear, expensive or fragmented. The model gets blamed — but the process was the failure point all along.
Read moreThe cost discipline framework
Cost work becomes effective when it is tied to the operating model instead of one-off budget cuts. A short framework for cost programs that survive the quarter.
Read more