A worked example of [[Shared Second Brain|building a shared second brain]] at organisational scale, traced through pain → cluster → quality bar → method → tool. Both questions matter: the pain-to-method line resolves to a cross-project decision log; the people-to-ownership line resolves to a standing review group.
## In one line
Teams repeating each other's mistakes across projects → cross-project decision rationale cluster, non-negotiables *discoverability* + *connectivity across projects* → [[Architecture Decision Record|shared technical decision log]] with a standing review group → [[Notion]].
^one-line
## The acts in one line each
Different teams kept making the same technical mistakes; per-project [[Architecture Decision Record|ADRs]] existed but were invisible across project boundaries.
^trail-line
Cross-project decision rationale - rare, high-consequence, stable, highly connected, consumed by future tech leads and by [[AI agent|agents]]. Non-negotiables: *discoverability across projects* and *connectivity across projects*.
^profile-line
Process-embedded capture as a shared [[Decision log|technical decision log]] common to the whole organisation, with a small standing review group as steward.
^method-line
## Full walkthrough
### Knowledge trail mapping
Different teams in the same organisation kept making the same technical mistakes. A new project would evaluate, say, an event bus pattern and reject it - then six months later, a different team would evaluate the same pattern from scratch, reject it for the same reasons, and never know the first team had been there. [[Architecture Decision Record|ADRs]] existed inside each project's own repo - invisible to anyone outside it. The trail went cold at the project boundary.
The people side: each project's tech lead held the rationale informally. Cross-project conversations happened in scattered Slack threads with no shared record. The informal brokers were a handful of senior engineers who happened to span multiple projects.
### Knowledge domain profiling
The cluster: decision rationale, cross-project, technical. Created by tech leads and senior engineers, infrequently. Consumed by future tech leads on different projects, by new engineers onboarding, and by [[AI agent|agents]] doing code review or architectural assistance. Volatility low - architectural decisions are stable. Connectivity high - each decision relates to others. Cost of being missing: high - repeated effort, repeated mistakes, contradictory patterns across the organisation.
Artifacts produced today: per-project ADRs (good). Cross-project index, shared search, shared linking: none (bad).
The quality bar:
- Discoverability - a tech lead on a new project must find a relevant prior decision in under five minutes, without knowing which project to look in
- Connectivity - decisions across projects must link to each other
- Maintainability - writing a new ADR must not slow the actual decision
- Survivability - must outlast any one project's lifetime
- Agent-readiness - structured and plain enough that an agent can pull context from it
Non-negotiables: *discoverability across projects* and *connectivity across projects*. Without those, the per-project ADRs already in place are not the failure; the failure is the lack of an organisational layer above them.
Steward candidates surfaced from the previous act: the senior engineers spanning multiple projects.
### Knowledge methodology matching
Process-embedded capture, specifically a shared [[Decision log|technical decision log]] common to the whole organisation. The discipline: every cross-project-relevant decision goes in the shared log, tagged with its project of origin, the alternatives considered, and the constraints that ruled. A small standing review group (drawn from the spanning senior engineers) confirms each entry so it counts as an organisational decision, not just a project one.
### Tool (illustrative, after the workshop)
Notion - linked databases for cross-project tagging and querying, comfortable for non-engineers to read, in the same workspace where the organisation's process documentation already lived. The tool followed the method (shared, queryable, cross-linkable), not the other way around.