How a use case gets in
Every entry must describe one problem and one mechanism. “AI for HR” does not qualify; “screen inbound CVs against a structured scorecard” does. Each entry carries at least one source describing a real deployment or a credible pilot with numbers, and every source is shown with its domain so you can weigh it yourself. Speculative pieces about what AI could theoretically do are rejected.
Where a claim is corroborated by several independent publications, we list them all. Independence is what counts, not volume — four restatements of one press release are one source, not four.
How scoring works
Three scores, each 1 to 5. They do not all run in the same direction: impact is better high, while implementation effort and data requirement are better low. That is why every score is shown as words — “Very high impact”, “Low effort”, “Advanced data needed” — rather than as a bare number you would have to decode.
| Dimension | 1 | 3 | 5 |
|---|---|---|---|
| Impact | Marginal convenience | Meaningful savings for one team | Changes a core P&L line |
| Implementation effort | Configure an existing tool, days | Custom workflow, 3–8 weeks | Multi-quarter build, org change |
| Data requirement | Works with documents any company has | Needs one clean system of record | Mature structured data across systems |
Where a score is genuinely between two values, we pick the lower impact and the higher effort — a directory that oversells is worth nothing.
A use case is marked a quick win when impact is 3 or above and effort is 2 or below.
Outcome numbers
Outcomes are published as ranges with sources, never as promises. Where a range cannot be sourced, the outcome is described qualitatively and no number is given.
Who maintains this
Maintained by the AI Use Cases Index editorial team. Every entry is researched against published sources, scored with the rubric above, and dated so you can see how current it is. Where a claim cannot be sourced, it is not published.
Found something wrong or out of date? The fastest fix is to tell us which page and which claim, and we will re-check it against its sources.