Client discovery

Phase 1 is where the governed loop runs hardest. The Phase 1 workspace (/engagements/{id}/phase1) has three tabs, in order.

1 · Problem framing

Define the client’s problem and objectives, and record success criteria. This is the entry point for every downstream AI feature — extraction, gap analysis, recommendations, and target design all read the framing for context, so a vague framing weakens everything after it.

You can attach framing sources (documents that describe the problem) and let the AI summarise them into draft objectives for you to edit.

2 · Discovery

• Artifact checklist — the list of material to request from the client.
• Upload — drop in what comes back. See Uploading artifacts for formats and what happens to each file.
• Jobs — each upload starts an extraction job; the panel shows progress and any warnings.
• Review queue — extracted candidates waiting for a person. This is the loop: see Reviewing AI candidates.
• Discovery plan status — whether you have enough material to move on.

3 · Capability

The capability model, built from what’s been approved. Sub-tabs run in dependency order:

1. Canvas — the capability hierarchy; edit structure, re-parent, add or retire capabilities.
2. Scoring — maturity scores. The AI suggests scores; a consultant reviews and applies them (the AI never writes a score itself).
3. Heat map — the scored model as a visual.
4. Current vs target — once Phase 3 has a target state, the two side by side.
5. App density — applications mapped onto capabilities.

There’s a parallel Process tab for business processes, including a BPMN editor for drawing or refining process flows.

The gate out of Phase 1

You advance when the capability and process models are in good enough shape and the discovery plan is satisfied. If you try to advance early, the gate tells you what’s missing.