Alliance AI

Agents / Model Risk Management

Intake Triage Agent

InternalAgent runMinimal riskGovernance · BU Restricted · 486 runs / 30d

Checks completeness and packaging conformance the moment a model, dataset or request is submitted, assigns the data classification and the model risk tier, picks the validation path and routes the case, so a validator never opens an incomplete file. Internal-data entitlements that need no discretion are granted here.

Governs the Alliance AI platform

Agent runAll governance agents →

An agent executes the process end to end and logs what it did. No human is in the path, because there is no discretion to exercise and no consequence to carry.

Acts for
Model Risk & Governance
Action class
write
Decisions / 30d
486
Safe-state returns / 30d
3

Processes covered

  • Team onboarding checks and role re-attestation
  • Intake triage and conformance
  • Data classification and model tiering
  • Internal-data entitlement grants
  • Entitlement expiry and renewal

Stages:Onboard & registerRequest & entitleBuild & validate

Memory

Stateless per submission; the routing record is the only artefact retained.

Safe state

Holds the case in 'needs review' and alerts the Model Risk queue owner if the schema validator or the classification rules are unavailable.

Reviewer time per case: 0 min with this agent, 12 without.

Ownership and sensitivity

Owned by

Group Risk ManagementModel Risk Management

Accountable owner: Dr. Kavitha Subramaniam

Who approves access

No approval needed. Self-service: subscribe and use, logged for chargeback.

Internal

Open to all staff. Aggregates, policy text, public series and synthetic data.

Self-service: subscribe and use, logged for chargeback.

Tools

  • Manifest schema validator
  • Data classification rules (Internal / Confidential / Strictly Confidential)
  • Model tiering rubric
  • Entitlement engine
  • Work-queue router

Data access scope

  • Read-only: submitted manifests, request forms and team onboarding records
  • Write: classification tag, model tier, routing decision and Internal-data entitlement grants
  • No access to training data or model artefacts beyond the manifest

Guardrails

  • Grants entitlements only on Internal assets to onboarded teams; anything Confidential or above is routed to a named approver
  • A non-conforming manifest is rejected with the failing field named, never patched silently
  • Classification and model tier can only be tightened automatically, never relaxed

Human-in-the-loop

  • A named approver decides every Confidential request the agent routes
  • The data owner decides every Strictly Confidential request

Orchestration

Trigger / IntakeLLM Orchestratorpolicy-rag-copilotManifest schema validatorData classification rules (…Model tiering rubricEntitlement engineHumansign-off

Every tool call is scoped by the declared data access above; the orchestrator cannot reach systems outside it. Owning business unit: Group Data & AI.

Live demo run

Watch the agent execute a real scenario step by step, every tool call, validation, and human checkpoint is traced and auditable. Typical run: ~4.8s.

Used in workflows

Retiring this asset would require these chains to be re-pointed first.

Audit

Updated 2026-07-29. Run traces retained 24 months for audit under the platform governance policy.

Community · 0 threads

Questions, findings and requests from the business units that use this asset. Owners reply here; threads with upvotes surface to the owning team's inbox.

No threads yet. Be the first to ask, or to share what you found.