Data Integration & Provenance

• Connectors and data servers exposing systems of record to agents
• Data contracts: schema, freshness and quality guarantees per source
• Lineage from source through transformation to the answer
• Entitlements: row and column security carried through to the agent
• Federated query across warehouses, lakes and operational stores
• Classification of sensitive fields before anything is indexed

Data Engineering, with Data Governance and Privacy

• Head of Data Engineering — accountable owner
• Data engineer — connectors, contracts, pipelines
• Data steward — source ownership and quality
• Privacy officer — classification, minimisation, retention
• Security engineer — entitlement propagation

Architectural recommendation, not a research finding.

Connect three sources properly — contract, lineage, entitlements — before connecting thirty carelessly. MCP data servers shorten integration work substantially, but the governance is the slow part, not the plumbing.

Signals to watch

• Freshness against contract per source
• Contract test pass rate
• Entitlement violations detected (target zero)
• Lineage coverage as a share of answers
• Time to honour a deletion request across indexes

1. The agent inherits broader entitlements than the human it acts for
2. PII copied into vector indexes where deletion obligations cannot be met
3. Stale or silently broken sources answering confidently
4. No lineage, so an incorrect answer cannot be traced to its source

Tool and standards snapshot · last reviewed approximately mid-2026