AI Portfolio — Index

The AI portfolio is the demand side of the transformation: what we build on the platform described in ../05-technical/_index.md, in what order, and under whose accountability. It has two registers:

1. ai-use-case-pipeline.md — every candidate use case, scored and prioritized. The funnel: idea → scored → pilot → production → retired. Nothing gets built that isn’t in the pipeline with a score and a KPI.
2. ai-agent-registry.md — every AI agent (assistants with tools and actions), with owner, approval points, risk level, and lifecycle state. Use cases describe demand; agent entries describe the running things and their governance.

How ideas enter

• Discovery interviews and department plans (../04-departments/it.md and siblings) — the initial ~20 below came from here.
• Department champions — standing agenda item in their monthly sync.
• AI helpdesk / feedback — thumbs-down themes and “I wish it could…” requests from the quality loop (../05-technical/rag-and-data-pipeline.md).

Anyone can raise an idea; the champion or program team drafts the pipeline row.

Scoring gate & approvals

• New ideas are scored (impact × feasibility, modifiers for complexity/risk — method in ai-use-case-pipeline.md) by the program team with the sponsoring department.
• Steering approves phase assignment, every pilot start, and every pilot→production promotion, per ../02-program/governance-model.md. Promotion requires the pilot KPI to have moved.
• Agents additionally clear the governance checklist in ai-agent-registry.md (owner, approval points, KPI) before pilot.

Both registers are living documents: update the row/entry in the same change as the decision that affects it.