Complete setup: all three pillars from zero

This is the full build-out: a fresh account to all three pillars running ( AI Spend, AI Governance and the engineering ledger). Each step says what it achieves and links to the doc that covers it in detail; nothing here repeats those pages. Done in order, the foundation takes a morning, the AI pillars a day or two of elapsed time (mostly waiting for data), and the ledger workflow its first full month.

Foundation

1. Create the organisation and connect your tools. The getting started guide covers the account, the dashboard checklist and the first sync. Connect an issue tracker and a code host: GitHub, Bitbucket, Jira.

2. Map every person's identities. Add people, link the accounts discovered during sync, and let email matching do the rest: people and identities explains why this single step decides the quality of everything downstream: effort, cost and AI attribution all resolve through it.

3. Set rates and organisation settings. Hourly rates give effort a cost; currency, timezone and the effort method are set once in Settings. The effort model explains the worklogs-versus-commit-sessions choice and what each claims.

AI Spend

1. Enable AI metering and connect the team. One settings change per person and every AI request is metered (cost, tokens, model, person) across Claude, GPT and OpenAI-compatible endpoints. The metering setup guide covers enablement, access URLs and the setup emails; the rollout guide covers the sequencing and the conversation with the team.

2. Let attribution bed in, then read the spend reports. Attribution ties sessions to tickets and initiatives and reports the untied remainder honestly; the AI Tax puts the spend next to the engineering cost it belongs with. Expect the first read to be a hygiene exercise.

AI Governance

1. Adopt the governance posture. Team-first privacy, correlational reporting and the audit trail are how the AI numbers stay defensible: the governance model is the reference, and it's worth sharing with the team before the first review.

2. Run a recurring governance review. Work the ranked exceptions, read the value story as labelled, record what was decided: the review scenario is the loop.

3. Add spend controls once the data is trusted. Budgets, team and person rules, and enforcement from alert-only up to blocking (budgets, limits and enforcement).

The Ledger

1. Classify the work. Classification rules turn tickets into categories: for capitalisation, the capitalisable/expensed split rests on rules someone can explain.

2. Group work into teams and initiatives. Teams and initiatives give reports their management-level shape, and initiatives carry AI spend beside labour cost, so the grouping pays off twice.

3. Run the monthly close. Confidence scoring, evidence gaps, readiness, lock, export: the R&D capitalisation scenario is the end-to-end workflow, and the stack-specific scenarios (GitHub + Jira, Bitbucket + Jira) cover the habits that keep the inputs clean.

When something doesn't look right

Empty reports, missing spend, work attributed to nobody, sessions without tickets: the troubleshooting page works through the usual causes in order. Most setup problems trace back to step 2: identity mapping is the foundation everything else stands on.