Baker Street
Baker Street combines hands-on AI and software delivery with the operating memory needed to keep that work understandable, testable, and reusable. It maintains this CDP Index as a shared evidence layer for CDP climate innovation projects: project context, source and repository references, decisions, open questions, evaluation evidence, operating notes, and reusable workflows.
The partnership model is designed to complement CDP's product, engineering, data, governance, and domain teams. Baker Street can investigate an opportunity, ship a bounded workflow, record how and why it works, and transfer the resulting knowledge so CDP can operate or extend it without restarting discovery.
Benefits of Baker Street as a partner
| Benefit | What it gives CDP |
|---|---|
| Faster starts with less re-discovery | Existing project, system, source, and decision context is carried forward through Index, reducing repeated investigation when a workflow changes hands or expands. |
| Delivery tied to evidence | Working software and workflow changes are accompanied by traceable sources, assumptions, decisions, limitations, and open questions. |
| Safer AI iteration | Evaluation cases, failure modes, human-review points, and fallback paths make AI-assisted workflows easier to assess and improve without treating model output as automatically correct. |
| Reusable capability across workflows | Patterns, connectors, evaluation checks, and operating notes from one project can inform later CDP workflows while preserving project-specific boundaries. |
| Durable handover | Runbooks, repository context, decision records, and maintainer guidance help CDP teams continue the work after a delivery phase ends. |
| Cross-functional coordination | A shared project memory gives product, engineering, data, governance, analytics, and domain teams a common evidence base for decisions and dependencies. |
| Partner-friendly delivery | Baker Street can work alongside CDP teams and existing platform, data, and delivery partners, making responsibilities, interfaces, and unresolved dependencies explicit. |
Partnership approach
- Investigate: map the workflow, users, systems, data boundaries, risks, owners, and evidence before selecting an AI or automation path.
- Deliver: build or improve a bounded workflow with observable outputs, human control, and practical acceptance evidence.
- Evaluate: capture expected behaviour, representative failures, limitations, and the checks needed before broader use.
- Record: update Index with the sources, decisions, implementation context, runbooks, and open questions needed for future work.
- Transfer and expand: pair with CDP maintainers, hand over reusable assets, and use real operating evidence to decide whether to stop, continue, or extend the pattern to another workflow.
Repository role
| Area | Description |
|---|---|
| Project memory | Durable, source-aware notes for CDP climate, AI, analytics, and data-platform work. |
| Project portfolio | projects/ connects project pages, implementation notes, operational evidence, and related partner context. |
| Repository intelligence | projects/cdp-repository-portfolio/ inventories relevant repositories and makes ownership, activity, and documentation gaps easier to inspect. |
| Delivery workflows | workflows/ stores reusable playbooks for technical discovery, AI augmentation, due diligence, migration, and delivery planning. |
| Partner and connector context | partners/ and connectors/ record stable responsibilities, integration surfaces, and operational dependencies. |
| AI-readable access | mcp/ exposes approved project memory and repository metadata to compatible local tooling. |
Current CDP scope
- Google Earth Engine and environmental disclosure data products.
- Adaptation & Action Explorer product, analytics, feedback, and operating context.
- CDP Ask AI delivery and evaluation context.
- AI-assisted corporate report and disclosure-question processing.
- Overture Maps and GERS ID adoption.
- Geospatial lookup and data-platform coordination.
- Repository intelligence, decision memory, evaluation evidence, risks, and open questions for CDP delivery work.
Working boundaries
- CDP retains ownership of its product, data, policy, licensing, governance, and external-communication decisions.
- AI-assisted outputs require the review and fallback appropriate to the workflow; this profile does not imply guaranteed accuracy or autonomous decision-making.
- Sensitive access details, personal relationship notes, mailbox evidence, pricing, and contract strategy remain outside this shared Index.
- New production integrations remain subject to the relevant CDP security, data-protection, platform, and operational approvals.