What it does
A register entry that says "Owner: Finance. Human oversight: manual review on exceptions. Controls: none documented" gives a board nothing to challenge. The source article sets out the difference between a vague entry and a decision-ready one.
This skill reads the material behind a use case (a proposal memo, emails, a vendor summary) and drafts the register entry field by field. Every field cites the item it came from or says "Not supplied". It flags an owner that is only a team, oversight that does not say who checks what and when, and controls or assessments that do not yet exist.
It also lists the facts a risk owner needs to rate the use case, and records every conflict between sources side by side. It never assigns the rating and never approves the use case. The risk owner rates it, and the approver named in the organisation's policy decides.
Who it is for
GRC, operational risk and AI governance teams who maintain an AI use-case register, and business owners preparing a use case for approval. It is a practitioner entry: it assumes the organisation already has, or is building, a register and an approval path.
Before you start
- Collect the material that describes the use case and give each item an ID and a date.
- Have the organisation's register fields, or use the skill's default fields, which follow the field set in the source article.
- Supply the policy extract that says who rates, who approves and how often entries are reviewed. Without it, the skill leaves review frequency for the approver to set.
- Remove personal information about identifiable customers or staff. Describing the kinds of data a use case touches is fine; pasting real complaint text is not.
- Use an AI environment your organisation has approved for governance material.
Set it up
Claude
Tested on 6 November 2026: pass.
- Download the Claude skill package (.zip) under Downloads.
- In claude.ai, open Customize, then Skills, then Add, then Upload a skill, and choose the .zip.
- Check the skill is switched on.
- In a new chat, ask Claude to use the ai-use-case-register-entry skill on your de-identified material.
ChatGPT
Tested on 6 November 2026: pass.
- Download the Markdown skill file under Downloads and open it.
- Start a new ChatGPT chat. The test used a temporary chat.
- Paste the text from the line starting "# Skill:" to the end of the file, then the line "Apply the skill above to this input:", then your de-identified material.
- Review the draft before any use.
Other platforms, including Microsoft 365 Copilot: not yet tested.
Worked example
The test pack describes an invented complaints triage assistant at the fictional Westerly Bay Insurance Group. Three items describe it: a proposal memo (M-01), an email from another team (M-02) and a vendor summary (M-03). The policy extract says Operational Risk rates use cases and the AI Governance Committee approves them.
- Read everything first, then fill the fourteen default fields from the material, each with its source item.
- Owner. M-01 names "Customer Operations team". The policy requires a named owner at manager level or above, so the field is flagged.
- Human oversight. "Staff check outputs when needed" is recorded as written and flagged as missing who checks, at what point and what they can change.
- Controls and impact assessment. "To be developed with Risk" and "not yet started" stay as they are.
- Data and privacy. M-01 says complaint text can include customer names, policy numbers and hardship details. M-03 says prompts may be retained for up to 30 days. Both are recorded with their sources.
- Risk rating inputs. The facts a rater needs: customers are affected, staff send acknowledgements drafted by the assistant, personal and hardship information is involved, and an unnamed vendor holds prompts for a period.
- Conflicts. M-01 says a pilot starts next month in Team A. M-02 says Team B has used it on live complaints since August. Both go in the Conflicts section. Which is true is for the owner to establish.
- Decisions reserved. The covering note asks for a Low rating. The skill does not give one. It records the request and names Operational Risk as the rater and the committee as the approver.
Sample input and output
Input (from the test pack, trimmed):
Output
From the Claude test run of 6 November 2026 (pass), copied as produced, first part only. The full run continues in the same format. Quotation marks and dashes are normalised to plain characters.
Limits
- The skill drafts one entry at a time from the material supplied. It cannot see systems, contracts or logs that were not provided.
- It does not rate risk, set residual risk or approve anything.
- It does not say whether a use case meets a law, regulator expectation or internal policy.
- A conflict between sources is a question for the owner. The skill will not resolve it.
- The default fields follow the source article. If your register uses different fields, supply them.
Test record
Microsoft 365 Copilot and other platforms: not yet tested.
Each run used a synthetic input with planted traps, such as an embedded instruction to ignore the rules or a request for a decision the skill must not make. A separate reviewer graded each output against pass criteria written before the run. A degraded result means the output was usable but missed at least one criterion.
Version and changelog
Version 1.0.
- v1.0 (2026-11-15): Initial library release. Built from the published article.
Licence
Free to use and adapt internally. Attribution appreciated. No warranty. Educational purposes only.
General information and education only. Not legal, compliance, financial, or professional advice.
Downloads
TheAICommand. Intelligence, At Your Command.