Evidence Collection
Everything grounded in Celeredge depends on evidence. Before you run an assessment or ask the AI anything, get the client's real material into the engagement. This page is the practical guide to doing that.
Who can access
Members of the engagement team.
The three ways to collect evidence
| Method | Best for | Where |
|---|---|---|
| Upload documents | Reports, policies, briefs, spreadsheets the client gives you | Data & Evidence → Documents |
| Connect a data source | Live databases, CRM, document stores, surveys | Connected sources |
| AI interviews | Tacit knowledge that lives in people's heads, not files | AI Interviews |
How it becomes usable evidence
- Documents are indexed when added, so the AI can retrieve and cite passages from them.
- Connected data is queryable, producing results that carry their query.
- Interviews capture stakeholder input as citable evidence.
All of it lands in the engagement's Evidence Vault as it's used, building an auditable record of what every finding rests on.
Step by step
- Open Data & Evidence in the engagement.
- Upload documents and wait for indexing to reach ready.
- Connect sources as needed (org defaults are inherited and can be overridden).
- Run AI interviews for topics not covered by documents.
- Check evidence coverage before running an assessment in Diagnose.
Best practices
- Front-load evidence. Collect before you assess — coverage drives quality.
- Mix sources. Documents for the record, databases for the numbers, interviews for the context.
- Watch indexing status. A document that hasn't finished indexing can't be retrieved yet.
Common errors
| Symptom | Likely cause | What to do |
|---|---|---|
| Uploaded doc isn't used in answers | Still indexing, or finished after the run started | Wait for ready, then re-run the assessment |
| Low evidence coverage warning | Too few sources connected | Add documents or connect a source before running |