All insights
Framework 10 min5 August 2026

How to rank AI opportunities by impact, feasibility, and risk

A practical scoring model for deciding what to automate first — and what to leave alone.

Ranking AI opportunities

Every leadership team has a list of places AI could help. The hard part is not generating ideas — it is choosing the first workflow that deserves budget, attention, and political capital.

This framework ranks candidates on four axes: impact, frequency, readiness, and risk. It is the same logic BYBO uses inside the AI Opportunity Blueprint — simplified so you can run a first pass internally.

Opportunity matrix · example scoring

Build first

High impact · Ready

Blueprint next

High impact · Gaps

Defer

Low frequency

Do not automate

High risk

Impact
Frequency
Readiness
Risk
Four quadrants — build first, blueprint next, defer, or do not automate.

The four scoring axes

Impact: how much time, revenue, quality, or visibility is lost when this workflow fails or slows down? Frequency: how often does it run — daily fire-fighting beats monthly edge cases. Readiness: do you have process knowledge, data access, and a willing owner? Risk: what happens if the system is wrong — customer harm, compliance exposure, or irreversible decisions?

Sample scores · 1–10 scale

Order exception recovery

Impact9
Ready7
Risk3

Policy Q&A for staff

Impact6
Ready8
Risk4

Full autonomous pricing

Impact8
Ready2
Risk9
Sample scores for three common workflow types — scale 1–10.

How to use the scores

Decision rules

  • Impact ≥ 7 and Readiness ≥ 6 → strong build candidate
  • Impact ≥ 7 and Readiness < 6 → Blueprint or process fix first
  • Risk ≥ 8 → human approval on every external action
  • Frequency < 4 → usually not worth automating yet
  • No owner → stop. Assign one before any vendor call.

The best first project is boring on a slide and expensive in real life — because that is where margin actually leaks.

Rank your top five workflows. Pick one. Run a two-week baseline: measure time, touches, and failure modes. Only then decide whether AI is the right tool — or whether a simpler integration would solve eighty percent of the problem.