Regulatory Radar

Recorded 2026-05-14. A 5-agent pipeline that watches a defined regulatory surface, flags material changes, maps them to the organization's controls, and drafts a governance-ready change brief — with a Governor pass that gates the output.

1. Step 1 of 5Scout

   ✓Completed

   Source discovery · 4.2s

   Input

   Configured regulatory surface: financial-services AI governance guidance across three illustrative jurisdictions, scoped to the prior 14 days.

   Output

   Identified 7 candidate updates across the surface; deduplicated mirrors and consultations to a shortlist of 3 substantive items for downstream analysis.

2. Step 2 of 5Analyst

   ✓Completed

   Materiality analysis · 6.1s

   Input

   The 3 shortlisted items plus the organization's AI use-case register and risk taxonomy.

   Output

   Classified 1 item as material (model-risk governance expectations for high-impact AI), 1 as informational, 1 as out-of-scope. Each carries a rationale and confidence note.

3. Step 3 of 5Mapper

   ✓Completed

   Control mapping · 5.3s

   Input

   The material item plus the organization's current control inventory (model inventory, validation cadence, human-in-the-loop policy, incident playbook).

   Output

   Mapped the change to 4 existing controls; identified 2 partial gaps — validation cadence for tier-1 models, and documented escalation thresholds — with suggested control updates.

4. Step 4 of 5Drafter

   ✓Completed

   Change brief · 7.4s

   Input

   Materiality verdict, control map, and gap list from the prior agents, plus the organization's standard governance brief template.

   Output

   Produced a one-page change brief: what changed, why it matters here, affected controls, recommended actions with owners, and a proposed timeline to next governance committee.

5. Step 5 of 5Governor

   !Flagged for review

   Pre-release review · 3.1s

   Input

   The draft change brief plus the project's release policy: cite every claim, no novel legal interpretation, escalate anything tier-1 to a human reviewer before publish.

   Output

   Approved for human review queue. Flagged one sentence as requiring a citation refinement and one recommendation as tier-1 — both routed to a named reviewer rather than auto-published.