AI and automation culture

Licences are bought quickly. Two months later a handful of people use the tool and the rest do not, and nobody knows exactly why. The question then asked is which tool would have been better, and that is the wrong question. Someone with no sense of which part of their own work is automatable will not automate it with better software either. The bottleneck is not the technology but the assumption that people will recognise by themselves what a computer could take off their hands.

Three things alongside each other, and none of them is a training course.

Working sessions where colleagues bring something from their own week and have it running by the end. Not watching a demo but producing something, because the conviction lands in the moment your own task suddenly runs itself.

A knowledge base that stays current. Most internal documentation is at its best on the day it ships. Keeping it up to date is therefore part of the work, not a job that is left over.

A recurring update showing what has become possible. What did not work last month works now, and without someone saying so the impression sticks at whatever happened the first time they tried.

The question of whether something can be automated now gets asked by people without a technical role, and that is the point where it starts working. What I deliberately do not put here is a percentage. The figures I have describe my employer's internal processes and they are not mine to publish.

My first round was about tooling. Showing what was possible, explained well, good turnout, and then nothing happened. The mistake was treating the technology as the interesting part when it is the easy part. What works is letting someone bring their own problem and wrestle with it until it runs. That scales worse and takes longer, and it is the only thing that sticks.