Impact ÷ Effort: How We Score Every AI Opportunity

Every opportunity in a Symphony Score carries two numbers: impact 1–5 and effort 1–5. Clients sometimes ask why we do not just estimate dollar returns. Because dollar estimates on unbuilt systems are fiction with a currency symbol — and fiction compounds badly when you rank twenty opportunities with it.

The impact scale

Impact scores against the business you actually described, not a hypothetical one:

5 — removes a named bottleneck. The thing the owner said out loud when we asked what hurts. Directly.

4 — frees meaningful weekly hours in a revenue-touching role, or measurably moves a conversion point.

3 — solid operational win. Hours saved, errors reduced, in a function that supports revenue rather than producing it.

2 — quality-of-life. Real, but the business would not notice its absence in the quarterly numbers.

1 — plausible on paper. We score it honestly and it usually should not be built yet.

The effort scale

Effort is about integration surface, not model sophistication:

1–2 — configuration territory. Proven pattern, clean data source, one system touched.

3 — one messy join. A legacy tool, an approval workflow, a data source that needs cleaning first.

4 — multiple systems, multiple stakeholders. Weeks, not days. Worth it when impact is 4+.

5 — organizational change disguised as software. The system is buildable; the workflow around it has to change for the value to land.

Impact divided by effort is not sophisticated. That is the point. A ranking you can argue with is worth ten rankings you have to take on faith.

Why transparency is the feature

Because both scales are published, a client can disagree — "that's a 2 for us, not a 4" — and the disagreement is productive. It surfaces context we missed. A black-box priority list produces nodding; a transparent one produces corrections. We would rather be corrected before the build than after.