The organisation is listening to you
When a senior leader talks about AI, people do not hear a technology update. They hear a signal about their jobs, their value and their future. Every phrase is read closely, and the phrases that sound harmless in a strategy meeting, like "efficiency" or "doing more with less", land very differently in a team that is already anxious.
Your managers handle the day-to-day conversations. The AI for Managers course prepares them for that. Your job is to set the message they are working within, and to make sure it is one they can repeat without flinching.
Fear about jobs
Some of your people are worried about their roles. Many will not say so. Pretending otherwise is visible and costly: the conversation moves to places you cannot see, with worse information than you could provide.
Name it yourself, and answer honestly. If you have no plans to reduce roles because of AI, say so clearly. If you genuinely do not know yet, say that, and say when you expect to know. If changes are likely, say what you will do: tell people early, explain the reasoning, and support those affected.
Never promise that no role will ever change. You cannot guarantee it, and one broken promise on this subject will undo trust in everything else you say. Promise what you control: honesty, early warning and fair treatment.
The honest message
A good message from the top is short and has four parts. What we are trying. Why we are trying it, framed as problems to solve rather than costs to cut. What will not change, including the standards and the human checks. And how people will hear about anything that affects them.
Test your draft by imagining it read aloud in a room of people who are worried. If it sounds like a press release, it will be treated like one.
Checkpoint
Name the job worry yourself, answer honestly including "I do not know yet", and promise only what you control: honesty, early warning and fair treatment.
Deskilling is a real risk
If AI drafts everything, people can lose the ability to do the work themselves, and with it the ability to spot when the AI is wrong. That matters most for junior staff, who learn judgement by doing the work badly first and being corrected.
Protect the skills you depend on. Keep some work done without AI, especially for people still learning. Make review a skill you train, not a box to tick. And ask your managers periodically whether people could still do the core work if the tools disappeared tomorrow.
Training that works
Training works when it is short, practical and done on real work. A session where people try a tool on a task they actually have, with someone who knows the pitfalls, beats a long presentation about possibilities. Budget for time, not just licences: people need space to practise, and that time should be visibly protected.
If you want outside help for larger teams, our partners page lists team training options.
Do not surveil
It can be tempting to measure adoption by tracking who uses which tools and how often. Resist it. Monitoring individual usage tells people that AI is something being done to them, and it produces usage theatre: people opening the tool to hit a number. Measure the work instead: speed, quality, errors caught. Where you need records for security or compliance, say openly what is recorded and why.
Where leaders should not use AI
This is the honest part, and it is worth saying to your whole organisation. There are places where AI is the wrong choice, even when it could technically do the job.
People decisions. Hiring, promotion, pay, performance, discipline and redundancy. These need a human who owns the reasoning and can explain it to the person affected. AI-assisted wording in these decisions blurs ownership exactly where trust matters most.
Anything you cannot explain. If you could not stand in front of the person affected and explain how the decision was reached, do not let a system make it or shape it heavily. A plausible explanation produced after the fact is not an explanation.
Messages that should come from you. Condolences, apologies, bad news and anything where the point is that a person took the time. People can tell, and it costs more than it saves.
Saying these limits out loud builds trust. It shows the organisation that AI is being used with judgement, not everywhere by default.
โ Weak prompt
Prompt
Write an all-staff message announcing our AI strategy.
Output
We are thrilled to embark on an exciting AI journey that will transform how we work, unlock efficiencies and empower every colleague to focus on higher-value work.
Every phrase an anxious reader will decode as a threat. It answers no real question and makes promises it cannot define.
โ Good prompt
Prompt
Draft a 200-word message from me to all staff. We are running three small AI pilots. Say what they are, that human checks stay in place, that we will not use AI for decisions about people's jobs, pay or performance, and that I have no plans to reduce roles because of these pilots. Say anyone affected by future changes will hear from me first. No words like journey, transform or empower.
Output
We are trying AI on three specific tasks: supplier query drafts, meeting summaries and first drafts of routine reports. Every output is still checked by a person. We will not use AI to make decisions about anyone's job, pay or performance.
Specific, honest about limits, and it names the line leadership will not cross. That is what people are actually listening for.
Here is a message I plan to send to staff about AI:
[paste it]
Read it as three different people: a worried junior employee, a sceptical
long-serving manager, and a union or staff representative.
For each, list:
- The phrases they will read as a threat or as spin
- The question they will ask that the message does not answer
Then suggest the smallest edits that would fix the worst problems.
Do not make the message longer than it is now.
๐ Quiz
Question 1 of 4What is the most credible thing a leader can promise staff about AI and jobs?