
AI and automation
AI Adoption Happens One Desk at a Time
Three training courses changed nothing. One hour on his own projects turned a sceptical 30-year veteran into the fastest AI user in the room. How it works.
A documented kill switch, a drifting AI quoting system and a year of losses. Why AI governance fails on culture, not on paperwork, and how to test yours.

There was a clear stop mechanism. On paper.
A mid-sized company had introduced an AI system for quote calculation. The managing director had announced the project at the company assembly as a great step forward; there was even a small press release. And there was, properly documented, a person with the authority to stop the system if something went wrong.
After a few weeks, a clerk noticed that the calculations for certain special machines were systematically too low. Not dramatically (a few percent), but reliably. She told her supervisor. He told nobody. Not out of malice, but because he understood exactly what it means to slow down a project the boss has just celebrated in front of the entire workforce.
The system ran for almost a year and produced a series of quotes that cost the company real money. The stop button worked the whole time. Nobody pressed it.
This is the most underrated part of AI governance. We write down who is allowed to stop, and we mistake that for control. Whether anyone actually stops is not decided by the documented right; it is decided by one question: does this organisation treat the person who speaks an uncomfortable truth as a troublemaker, or as someone doing their job? And that answer is fixed long before the first AI project starts.
Think of the last genuinely bad news that made it to the top of your company. Who delivered it, and what happened to that person afterwards? Were they heard and backed, or were they quietly labelled “difficult”? That one memory says more about your AI governance than any policy document. An organisation that punishes the messenger today will not stop an AI tomorrow, however cleanly the stop right is documented.
Want to know whether your organisation would press the button in time? Book a demo or try our ROI calculator.
Read on

AI and automation
Three training courses changed nothing. One hour on his own projects turned a sceptical 30-year veteran into the fastest AI user in the room. How it works.

AI and automation
The same AI tool makes a strong engineer faster and a weak process worse. Why the expert in the loop decides whether AI output is an asset or a liability.

AI and automation
Most AI business cases are built on saved positions, and most disappoint. How to measure cycle times, error costs and freed capacity honestly instead.