Skip to main content
LLM Squared

AI scoping records

Where does AI support your institution?

Start with the places AI may already support work, including features inside software your institution owns.

What examiners look for

The Work Program asks examiners to look beyond formal models and public applications. Operations, compliance, fraud, service channels and staff support can all matter. A feature may use AI even when its product name never mentions AI. A list limited to approved chat tools can therefore miss useful context.

CSBS Examiner Work Program, initial scoping 1. Checked October 16, 2026.

Which records answer the question

  • Keep the AI inventory alongside the business areas reviewed.
  • Retain the vendor-product list, including possible embedded features.
  • Link recent changes to the product and business owner.
  • Use staff guidance to distinguish approved tools from tools found during interviews.

See all 19 requested records.

Where the answer directs review

A Yes can point examiners toward Governance and Oversight, plus AI Inventory and Use Cases. A No or unclear answer may prompt a cross-check. The guide names vendor lists, software lists, approved tools and recent platform changes. Explain the checks made before reaching a conclusion about use.

Core Examiner Guide, page 3. Checked October 16, 2026.

Two gaps to check

The software list names products, but nobody checked their AI features. Ask product owners what changed and request feature-specific records. Keep unknown capabilities open instead of assuming an older purchase contains no AI.

Staff say AI is absent because no chatbot was bought. Ask about drafting, fraud review, analytics and internal support. Record the areas checked and the limits of that work before closing the inventory question.

How to prepare the record

Choose the business areas covered by the review. Ask each owner about current tools, new releases and informal use. Keep the date, role and material behind each answer. An interview gives a starting point; product records may confirm or challenge it. Separate a tool found during the review from a use that the institution has accepted.

A row behind the question

Illustrative inputs, not a client.

An illustrative evidence row
UseWhere it runsData it can touchOwnerEvidence on handQuestion still open
Feature inside an existing productVendor serviceNot yet identifiedProduct owner role not assignedProduct proposalDoes the feature use AI, and is the feature active?

What the reviewer should retain

The useful result is a bounded coverage record. It shows where the institution looked, what was found and what still needs checking. Keep areas outside that scope visible. A later product release can reopen the question even if an earlier review found no AI.

Prepare the next step

Continue with the inventory question to turn findings into a usable register.