What the organisations that make AI stick do differently. Ten principles you can test on Monday, drawn from how banks, consultancies, pharmaceutical companies, governments and international organisations have adopted AI, and from what earlier transitions teach.
Nearly every organisation I meet has given its people access to an AI assistant. Far fewer can point to a piece of work that has changed because of it. The organisations that make AI stick do the same handful of things, whatever their sector, and they do them together. The failures are almost never the technology. They are management decisions.
These are the ten things. Each one comes with a question to ask of your own organisation. The questions you cannot answer are your blockers.
One senior person answers for an outcome. Not a committee, not IT alone. The owner secures the resources, and the team leaders under them run the practice that changes behaviour.
A sanctioned tool with clear data rules, quicker to reach than a personal chatbot. People use AI whether or not you have decided they may, so the choice is between a tool you govern and one you cannot see.
A short course on potential, limits and rules for all, then training on your own work. Of the ten practices, this is the one with the strongest evidence behind it.
Volunteers with protected hours, connected to a central point that turns their finds into shared assets. Without the second half, use stays hidden in one person's browser and never scales.
Pick the tasks the model does well and change the steps, roles and handoffs around them. If no step, role or handoff has changed, you have a faster version of the same process.
Someone checks what leaves, against something, and a quality measure sits beside every usage figure. Almost every public reversal I have studied came from a missing check, not from the model.
Decide what will count as success, measure it first, and keep a comparison. Without a baseline every later claim is a story, and self-reported gains overstate measured ones.
Let many experiments start, fund the few that show value, retire the rest. The organisations that get it right treat adoption as a portfolio, and that includes retiring their own builds.
Nobody is rated on how much they use the tool, and staff are heard before decisions that touch jobs. Where usage became a target, the correction came from outside: a union, a tribunal, a public backlash.
Stage access, rehearse the risks, let readiness set the date, tell people the limits. The fastest way to lose people is to promise more than the tool delivers.
The point is not whether your people have access. The point is whether the work has changed.
Each principle in full: what it is, what it looks like in practice, where it goes wrong, and the one question to ask on Monday. Plus the ten questions on one page, as a self-assessment for your leadership team, and the sources behind every claim.
I use your address to send you the document and, now and then, a note on AI adoption. Nothing else, and a reply to any of them stops the notes.
The document opens in a new tab. If it does not, use the button.
Download the PDFThe principles describe what the organisations that get it right do. The AI adoption plan is how I help you do it, over six months, so that your own people carry it on without me:
The executive session and the small group coaching are the two other ways of working together, described on the main page.
A 30-minute call to go through the ten questions with you, and to see which of the three formats fits.
Schedule a call