Notes from the work
AI at work. Questions worth asking. What we are learning.
Featured publication
SubstackNote
Our new walkthrough follows an employee describing their AI use, its review against company rules, and the human decision that contributes to the company view.
Read on Substack
Writing and discussion
9 publications
A practical discussion of the first response to employee AI disclosure: a named decision owner, what can continue or must pause, a review date and an alternative that can do the job.
An aviation reporting analogy opens a six-post discussion of shadow AI, safe disclosure and the useful human answer employees should receive.
A discussion about employee AI disclosure, trust, review ownership and what happens after an amnesty. Invites implementation experiences, results and objections, with background from Mario Thomas's work.
Notes from the AWS and SANS cloud AI session. Three vendor demos illustrate identity, data and agent controls; permission to read data does not automatically include permission to disclose it.
A dated development test shows how assigning one passage ID to only one case can stop two AI systems from each citing their evidence. This is a source-contract observation, not a proven model failure or repair.
Our new walkthrough follows an employee describing their AI use, its review against company rules, and the human decision that contributes to the company view.
A product demonstration of rulebook review and recorded human decisions. The company and every figure shown are fictional.
A note from Andres at Andes on the information a reviewer needs. The linked film is an illustrative animation, not a compliance score.
What does it do, who does it affect, and what is still unknown? Our earlier illustrative animation keeps sources and open questions visible. People make the decision.
Follow Andes where you read, write and ask questions.
Product profiles on independent software directories. These links are not ratings or endorsements.