Team tool · Template
Verification levels
Agree which outputs get a glance, which get spot checks and which get full verification, so everyone checks in proportion.
Principles involved
Overview
Nobody can check everything. Agreed verification levels tell everyone how much checking each kind of output needs before it leaves their hands.
The three levels
| Level | For | What you do |
|---|---|---|
| A glance | Internal drafts, ideas, notes to yourself | Read it through. Does it answer the question you asked? |
| Spot checks | Anything you share with colleagues | Check the names, dates, figures and claims that matter most. |
| Full verification | Anything published, signed, sent to a client or used to decide | Open every source, recalculate, confirm quotes and dates, and ask a second person when the stakes are high. |
What full verification means
- 01Open every source and find the sentence that supports each claim.
- 02Recalculate figures from the original data.
- 03Confirm quotes, names, dates and amounts.
- 04Check that the result fits your country, market and client.
- 05Ask a second person when the stakes are high.
The template
Each field says what to record. The fictional example shows what a finished record can look like.
Fictional example: the marketing team at Lumen, an invented software company.
- 01Output
-
Name a kind of output the team produces with AI.
Fictional example
Customer case study for the website.
- 02Level
-
Choose a glance, spot checks or full verification.
Fictional example
Full verification.
- 03What to check
-
List the checks this output needs.
Fictional example
Every quote approved by the customer in writing, every figure traced to the customer’s data, and product claims checked with product marketing.
- 04Who checks
-
Say who does the check, and who is the second reader if one is needed.
Fictional example
The author, then the content lead.
- 05What we keep
-
Say what evidence of the check you keep, and where.
Fictional example
The customer’s written approval and the source data, in the case study folder.
Sources
What each source establishes, and its limits. The practices and recommendations on this page are ours, and the facts come from the sources. See every source we use.
- Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile (NIST AI 600-1) US National Institute of Standards and Technology · July 26, 2024 · Voluntary framework Lists twelve risks of generative AI, including confabulation (confidently stated false content, sometimes with invented citations) and human-AI configuration, which covers automation bias and over-reliance. Limits: Voluntary. It was issued under an executive order that has since been revoked, and the wider framework is under revision.
- Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools Journal of Empirical Legal Studies (Stanford RegLab and HAI researchers) · April 23, 2025 · Peer-reviewed research On 202 legal questions, leading AI legal research tools hallucinated on 17% to 33% of queries, and a general-purpose model on 43%. Limits: A snapshot of tools tested in 2024. The tools have changed since.