AI capability is not AI authority. Most companies acquired the first without ever defining the second.
That is the argument of a paper I have just published together with Amanda Cunningham, founder of Savvy Pixel.
Her half covers what happens inside the organisation. AI never formally joins a company. Nobody interviews it, nobody signs anything. It arrives through accounts, features and updates, and one day it is talking to your customers. Her fix: induct AI the way you would induct a new employee, with eight questions leadership must be able to answer. Purpose, context, access, authority, boundaries, accountability, evidence, review.
My half starts where hers ends: the moment the AI meets a customer. From 2 August 2026, Article 50 of the EU AI Act requires telling people they are talking to an AI. That is the legal floor, not the customer proposition. A customer has three real questions: how does this company use AI at all, what can this particular bot do, and how do I prove my case when something goes wrong. Three documents answer them: an AI policy, an AI service passport, an AI receipt. And a periodic test purchase checks that the answers are still true.
Induction establishes the promise inside the organisation. Accountability tests whether that promise survives contact with the real world.
The paper is 12 pages, free, no registration: From AI Induction to AI Accountability, first in the library.