Goal, baselines, and deliverables

A project can end in a document: an RFP for a replacement, a change request to a vendor, a new process draft, or a deck. Your AI tool writes it. The workspace holds the evidence it rests on and records what was produced.

All of this is on the project’s Outcome tab, in the browser and in the desktop app.

Set the goal

1. Open the project and select Outcome.
2. Select Set goal.
3. Enter the Goal: what the project is working toward and the decision it feeds.
4. Fill in what you know:
   • Deliverable: what must be handed over. Leave it empty if you don’t know yet; your AI tool can suggest a form.
   • Audience and Decided by: roles, not names.
   • Due.
   • Constraints: limits and non-goals the deliverable must respect.
5. Select Save goal.

The discovery skill also asks what the work is for and saves the goal once you have said it.

Freeze a baseline

A baseline is a frozen copy of the goal, the map, the decisions, and the session results at one moment. The deliverable is built from a baseline, so it keeps pointing at the same evidence while the map stays editable.

1. On the Outcome tab, select Freeze baseline.
2. Optionally enter a Label, for example Before the steering committee.
3. Select Freeze baseline.

Each baseline shows:

• Not final: what was unsettled when it was frozen, such as open questions, unconfirmed proposals, or a session that was still open. This never stops a freeze.
• How much the map has changed since, for example 4 changes since: 2 items, 1 decision.
• Findings: the frozen evidence as one document. Download it as Markdown, or download the raw records as JSON.

Only a person can freeze a baseline. Your AI tool asks you to when there is none or when the map has changed since the last one.

Produce the deliverable

Ask your AI tool to conclude the project:

/process-discovery:conclude "Quotation tool"

If the desktop app installed the kit, use /conclude with the same argument. Add what to produce if the goal does not say, for example "Quotation tool" RFP for a replacement.

The tool reads the goal and the latest baseline and chooses the form. It writes the files into a folder in its working directory, outcomes/<name>/r1/. It then registers the revision: the baseline it was built from, and each file’s name, size, and checksum. The workspace keeps this record, not the files.

To revise a deliverable, ask again. The tool writes r2 next to r1 and leaves the earlier revision as it is.

Approve a revision

The Deliverables section lists each deliverable with its revisions and the baseline each was built from.

1. Select Approve on the newest revision.
2. Confirm.

A later revision needs its own approval. Only a person can approve; your AI tool cannot.

The badges above the list show how far the project has got: Goal, Baseline, Deliverable, and Approved. If the map has changed since the baseline the newest revision was built from, another badge says by how much.

Next

Reference lists the MCP tools your AI tool uses for these steps.