The Objection

Why AI training doesn't stick

Daniel Kerambrum Bicourt · Founder, CO-ORDI · July 2026 · 4 min read

The pattern is so consistent you can set a calendar by it. The workshop lands well. Everyone leaves energised, three people install new tools that afternoon, someone writes "AI champions" in a Slack channel name. Two weeks later, the business runs exactly as it did before.

Why the fade happens every time

• Knowledge isn't capability. Knowing what a prompt is doesn't build the reporting pipeline. The distance between "understands AI" and "ships working systems" is months of practice your team doesn't have spare.
• Nobody owns it. Training distributes responsibility across everyone, which in practice means no one. The moment client work gets busy, the AI experiments are the first thing dropped.
• The field moves monthly. Whatever was taught in the workshop is partly out of date within a quarter. Without someone whose job is to keep up, the knowledge decays even if the enthusiasm survives.
• Systems beat sessions. Behaviour changes when the default way of working changes. A workshop changes nothing about the default; a live system your team uses every morning does.

What actually sticks

Capability sticks when two things are true: working systems exist inside the business, and someone is accountable for improving them. That is the entire logic of the embedded expert model. The expert doesn't teach your team to build pipelines. They build the pipelines, run them, and improve them monthly, while your team keeps doing the jobs they were hired for, now with the manual work draining away.

Your team still learns, but the way people actually learn at work: by using live systems daily and sitting next to the person who builds them. Induction and documentation come with every CO-ORDI build for exactly that reason.

Where training does earn its keep

Paired with an operator, training works. Once systems are live and owned, a focused session on how to use them well multiplies the return. The order matters: systems first, training second. Run the other way round, you're funding enthusiasm with nothing for it to land on.