How to run an AI use-case discovery workshop for your enterprise

Most enterprise AI programs stall because leadership picks tools before picking problems. A use-case discovery workshop flips that order: you map where reasoning across your own knowledge would change a decision, then you wire the tooling to fit. Here's how to run one in a week.

Step 1: Inventory the decisions, not the tasks

Spend the first session listing the ten decisions your team makes each quarter that carry the most weight — pricing calls, hiring bets, roadmap cuts, clinical protocols, legal risk memos. Ignore tasks that just need to move faster. You're looking for decisions where more context would change the outcome.

Step 2: Trace the context that already exists

For each decision, list the sources someone would need to read to make it well: customer calls, past decks, PDFs, Slack threads, screenshots, voice notes from the field. If that context lives in ten places and no one reads all of it, you've found a use case.

Step 3: Score by ROI, not by novelty

Rank candidates on three axes:

• Decision frequency — how often is it made?
• Cost of a wrong answer — dollars, hours, risk.
• Context gap — how much of the needed knowledge is scattered today?

Pick the top three. Everything else waits.

Step 4: Prototype against a real knowledge graph

Don't prototype with a demo dataset. Ingest the actual sources for one decision into an enterprise knowledge graph and let a small group run real queries against it for two weeks. Measure whether the answers cite the right sources and whether the decision changes.

Step 5: Set the rails before you scale

Before rolling out, define who can see what, which reasoning mode is used for which query type, and how answers get audited. Then hand it to the next team.

Skip the workshop and you'll end up optimizing yesterday's work instead of unlocking new decisions.