AI Banking Resources · Template

AI Use-Case Inventory

A register for documenting every approved, restricted, and proposed AI use case before it becomes normal work.

For: Compliance, risk, operations, and AI program owners6 min

Template preview

AI Use-Case Inventory

A register for documenting every approved, restricted, and proposed AI use case before it becomes normal work.

01

Use case

  • Business purpose: the operational, member, risk, or compliance reason this use case exists.
  • Owning department and named owner.
  • Current status: proposed, sandbox, approved, restricted, retired.
02

Tool and vendor

  • Tool name, vendor, and version if known.
  • Approved-list status and approval date.
  • Vendor agreement, retention setting, and review owner.
03

Data class

  • Public, internal, confidential, NPI, regulated, or examination-sensitive.
  • Whether customer-identifying data is prohibited, restricted, or allowed only in an approved private tool.
  • Sanitization rule before any prompt is used.
04

Human review

  • Reviewer role and backup reviewer.
  • Review criteria: accuracy, data handling, regulatory references, tone, and final-use approval.
  • Evidence retained: prompt, output, reviewer notation, or ticket reference.

Use case

Name the AI-assisted work in plain language. The title should tell a reviewer what staff are trying to accomplish without needing a demo.

  • Business purpose: the operational, member, risk, or compliance reason this use case exists.
  • Owning department and named owner.
  • Current status: proposed, sandbox, approved, restricted, retired.

Tool and vendor

Record the exact tool being used, including whether it is a public AI service, a vendor feature, or a private deployment.

  • Tool name, vendor, and version if known.
  • Approved-list status and approval date.
  • Vendor agreement, retention setting, and review owner.

Data class

Document the highest-risk data that may enter the workflow. If the workflow can operate without confidential or regulated data, say so explicitly.

  • Public, internal, confidential, NPI, regulated, or examination-sensitive.
  • Whether customer-identifying data is prohibited, restricted, or allowed only in an approved private tool.
  • Sanitization rule before any prompt is used.

Human review

Every reusable AI workflow needs a named review point before output is relied on, sent, filed, or shared.

  • Reviewer role and backup reviewer.
  • Review criteria: accuracy, data handling, regulatory references, tone, and final-use approval.
  • Evidence retained: prompt, output, reviewer notation, or ticket reference.

Risk tier and cadence

Assign a risk tier and a review cadence so the inventory remains useful after the first approval conversation.

  • Low, medium, high, or blocked with reason.
  • Re-review trigger: tool change, policy change, incident, vendor update, or annual cycle.
  • Next review date and accountable owner.
Adapt before adopting

These are starters — not final policy.

Every template names a section your institution should change. Bring it to your committee, your auditor, and your examiner before adoption.

AI Use-Case Inventory — AI Banking Resources — The AI Banking Institute