An anonymous real example
During a bid-team exercise, an AI assistant supplied a quotation and a source. The participant challenged it: were those words really in that document? The assistant confidently said they were. Opening the source showed that the quotation was not there. Further challenge eventually exposed the mismatch. The conversation had supplied reassurance before the document supplied the answer.
The assumption worth revisiting.
In the reported exercise, returning to the original source exposed a claim that the assistant had asserted. Now consider the organisational implication: if a proposal and its reviews all reuse the same unsupported assumption, several approvals may add less challenge than they appear to. That wider propagation is a scenario to investigate, not a reported incident across those functions.
What new evidence did each of our approvals add?
Give the capacity a purpose.
Trace the basis for a decision across functions and make time for an independent perspective where the consequence warrants it. Ask what genuinely new evidence each review contributes.
How this fits the capacity allocation →
Use AI to challenge your first interpretation
Explore this question without assuming our first interpretation is right: What did the check add, other than another expression of confidence? Offer competing explanations, identify assumptions and suggest evidence we could obtain from the people doing the work.
Treat the answers as possibilities to test with people and evidence.
Go deeper into the evidence.
Judiciary of England and Wales · Published judgment
The judgment addresses false legal citations and the responsibility of legal professionals to verify research and authorities.
Specific proceedings, not a general error rate for AI legal tools. The practical application here is a review workflow, not advice on a particular dispute.
ACM CHI 2025; Microsoft Research · Peer-reviewed survey
319 knowledge workers supplied 936 examples of AI use. Greater confidence in AI was associated with less reported critical-thinking effort; confidence in one's own ability was associated with more. Thinking shifted toward checking information, integrating output and overseeing the task.
Self-reported correlations. This does not prove AI causes declining intelligence or long-term skill loss.
video
Introduction to the NIST AI Risk Management Framework (AI RMF 1.0): An Explainer Video
NIST
Gives leaders a shared starting point for deciding who owns AI risks and what needs checking throughout a system’s use.
Listen or watch for: Who owns the risk after a tool moves from pilot to everyday use?
About this resource
An introduction to a voluntary US framework, not UK legal advice or a certification. NIST says the framework is being revised.