Point One Zero Platform architecturev11 · generated viewSign in to edit
Playbook › Track 3: Data & integration

Track 3: Data & integration

Track 3: data, systems, and integration. Led by data engineering. This track maps everything that exists, decides where truth lives, and turns every connection into a specification that lands as parameters, so integration is configured, never improvised.

Step by step

Inventory

  1. Full systems inventory. Systems of record, document stores and VDR-class restricted stores, data warehouses, insight and reporting systems, existing vector databases, and any ongoing AI implementations, gathered through Discovery section D with owners and admin contacts.
  2. Document and data needs per capability. For each roadmap item from Track 2: which documents, which tables, which fields, and how fresh the data must be.
  3. Authority mapping. Where truth lives for each entity; inputs to the source authority weighting policy; restricted stores flagged for entitlement-checked retrieval only.

Specify the south-facing layer

  1. South-facing MCP specifications. One tool server per source, tool definitions written to the published tool-design standard, permission scopes defined once per server. These specifications populate the source catalog agents consult at runtime.
  2. Reusable agent match. Compare capability needs against the P1Z agent register: reuse as is, reuse with customization (selected on the plane, never forked), or new build through the ADLC. Every step of every candidate passes the determination rule so the deterministic coding layer and the agent coding layer are separated before development starts.
  3. Integration requirements, parametrized. Every connection is classified by pattern: ETL batch, integration hub, API, real-time, or near-real-time, and documented in the format that lands directly as plane values: source selection, credentials path, data contract, field mappings, sync mode and schedule (the C5 band). A requirement that cannot be expressed as parameters goes back for rework.
  4. Existing AI implementations. For each one found: integrate, wrap behind a tool server, or retire, with the decision recorded on the Decision Register.

Deliverables

DeliverableLands as
Systems inventory with ownersSource-estate value (C6) and the source catalog content
Authority mappingInputs to the authority weighting policy in Knowledge
South-facing MCP specificationsTool server definitions, one per source
Integration requirement sheetsParametrized; filled into the C5 band
Reuse mapReuse / customize / build decision per capability

This track uses: Discovery section D, the C5 and C6 bands, the Authoring contract page for tool-definition standards, and the glossary entries for source catalog, restricted store, and canonical record.