Verifying AI output
5articles
November 2026
5 articles on this page
AI for leaders11 min readLawyers, ChatGPT and the cases that never existed
For managers and owners who sign documents drafted by other people or by AI. A New York federal court fined lawyers USD 5,000 over invented case law; its finding of bad faith turned on how they responded once challenged.
- AI for leaders10 min read
Which tasks should you give AI? A stake-and-ambiguity matrix
For team leads, operations managers and SME owners listing the week's tasks and unsure which to hand to AI. Score each task on the damage if it is wrong and on ambiguity, read off one of four cells, then run a single trial before you trust the cell.
- AI for leaders10 min read
The five-rung ladder: check an AI draft before you sign
For managers and owners about to sign or send a draft written by AI. You get five checks in order, who does which and a rough time cost, worked through an invented food distributor in Can Tho.
- AI for leaders10 min read
When an airline's chatbot gave wrong advice, the airline paid
For owners and managers putting a chatbot in front of customers. A tribunal in Canada held Air Canada responsible for what its chatbot said; this post reads the decision closely and suggests six changes to your process.
AI for leaders9 min readHallucination, sycophancy and recency lag: three AI errors to catch
For owners and managers who use an AI assistant to draft reports, challenge plans or look up rules. You get three tests, each built to take about 20 minutes, to check your own tool for invented sources, flattery and out-of-date answers.