Public docs

Engagement Process

The engagement process is intentionally short and evidence-led. It produces controlled artefacts that a delivery team can inspect, challenge, and retain.

1. Frame the decision

Agree the model area, delivery question, accountable reviewers, relevant systems, and what must not be changed. A pilot usually starts with one domain rather than an enterprise-wide inventory.

2. Establish the baseline

Collect approved-to-share inputs such as mapping extracts, model notes, validation reports, datasets, and existing decisions. Represent authoritative model knowledge as canonical files; keep raw inputs and generated outputs distinct.

3. Validate and investigate

Run deterministic validation, profile evidence, build derived indexes, detect gaps, and trace selected fields or relationships. Findings must retain their evidence and detection mode.

4. Review controlled options

Review gaps, impact, decisions, and PatchProposals with accountable people. AI can help draft a proposal; it cannot silently update canonical truth. High-risk changes remain behind approval gates.

5. Hand over the working model

Deliver the canonical baseline, generated reports, reviewable proposal/decision trail, and the repeatable commands or local workflow needed to continue. Generated indexes are disposable; canonical files stay the source of truth.

Get started

The best first message includes the delivery problem, a candidate model slice, systems involved, and whether representative inputs can be shared under the appropriate controls. Continue at Contact.