Guide
How to prioritize AI projects
Prioritize AI projects by scoring each candidate on value (how much time, money or outcome it could move) and ease (data readiness, judgment required, implementation complexity and evidence), then start with the ones that are high on both. A smaller, ready opportunity usually beats a larger one whose data or process isn't ready.
Updated
Score value and ease separately
Combining everything into one score hides the trade-off you need to see. Keep two axes. Value asks how much changes if this works. Ease asks how likely it is to work soon.
- Value inputs: frequency, people involved, hours or cost today, volume and customer value for growth cases.
- Ease inputs: are the inputs digital and structured, how much expert judgment the output needs, how many systems it touches, and how strong your evidence is.
Read the four quadrants
- High value, easier: do first.
- High value, harder: plan. Fix data or process readiness first.
- Lower value, easier: quick wins, only if they build capability.
- Lower value, harder: skip for now.
Compare across different kinds of value honestly
A capacity figure and a revenue scenario are different kinds of number. Rank them as an ordered list of things to investigate, and never add them into one headline.
When AI is not worth it
- Prioritizing by vendor demos instead of your own volumes.
- Scoring with made-up benchmarks rather than measured hours or counts.
- Starting with the hardest, most visible project to prove a point.
Compare your own opportunities
The free Opportunity Map ranks what you enter on value and ease, with every assumption shown. About 6 minutes, no signup.
Questions
- How many AI projects should we start with?
- One pilot, with one or two others planned. Parallel pilots dilute ownership and make it hard to tell what caused a change.
- What if we have no data on hours or volumes?
- Estimate ranges and mark them as low confidence. The first step for that candidate becomes a two-week measurement, not a build.