Idea Mining
This playbook helps run a session for collecting, structuring, and triaging AI ideas. The output should not be a board of wishes; it should be a set of initiative candidates with owners, impact hypotheses, and next steps.
When to use
- launching an AI office or a new domain;
- forming an initiative portfolio;
- after employee training;
- when a business unit has many scattered requests;
- when looking for scenarios for a specific AI product.
Participants
| Role | Purpose |
|---|---|
| Facilitator | Keeps the session structured and turns discussion into cards |
| Process owner | Explains current work, pains, and constraints |
| AI office | Checks whether ideas fit the conveyor |
| AI product owner | Suggests possible delivery route |
| Data / architecture / security | Flags blockers and requirements |
| Finance or business control | Helps shape the impact hypothesis |
Session steps
- Define a specific goal.
- Break down the current process.
- Generate ideas by scenario type: knowledge search, drafting, classification, document checks, prediction, automation, employee assistance.
- Triage each idea by problem, impact, owner, data, product, risk, and next step.
- Select initiative candidates.
Candidate criteria
An idea can become an initiative candidate when:
- there is a business owner;
- the process problem is clear;
- an impact hypothesis can be stated;
- no obvious duplicate exists;
- no immediate data or risk prohibition exists;
- the next validation step is clear.
Output
- 3-10 initiative candidates;
- postponed idea list;
- owners for clarification;
- tasks for data, impact, product, and risk checks;
- date of the next triage or gate discussion.
Summary template
Idea:
Process:
Problem:
Current metric:
Expected impact:
Business owner:
Potential AI product:
Data:
Risks:
Similar initiatives:
Next step: