In many organisations, AI capability lives in one or two people. They know which prompts work, how the workflows fit together and who to ask. When they leave, the capability leaves too. Stopping AI skills walking out the door takes two things: making the capability less dependent on individuals, and giving AI-skilled people good reasons to stay.
Why this risk is growing
People who are good at applying AI at work are increasingly valuable. Other employers notice them. At the same time, AI skills are often informal and undocumented, built up through personal experimentation. That combination makes AI capability unusually fragile.
Part 1: Make capability less dependent on individuals
Spread skills across the team
Train several people on each important AI workflow, not just one. Pair experienced users with newer ones.
Document what works
- A shared prompt library for common tasks.
- Written workflows showing each step and where humans review.
- Notes on what does not work and why.
Build AI into standard processes
When AI use is part of the official process, taught to new starters and written into procedures, it survives individual departures.
Develop multiple champions
One champion per team is better than one champion for the whole organisation.
Part 2: Give AI-skilled people reasons to stay
Recognise the contribution
Make AI expertise visible: credit people who build useful workflows, include AI capability in performance conversations and give champions formal recognition.
Offer growth
People who build AI skills want to use them. Give them bigger problems, more responsibility for AI initiatives, or a pathway into leading AI adoption.
Invest in their development
Funding further training signals that you value their skills and want them to grow with you. Many people leave because they feel their development has stalled.
Reward fairly
If AI skills significantly increase someone's value, review whether their role and pay reflect it. Otherwise, the market will do it for you.
A quick risk check
Ask about each important AI-supported workflow:
- How many people could run it tomorrow if one person left?
- Is it written down?
- Is it part of how new starters are trained?
If the answers are "one", "no" and "no", that workflow is at risk.
The manager's role
Managers are central to both parts: spreading skills across their teams and recognising the people who drive AI adoption. Our Leading AI Innovation & Organisational Transformation course covers appointing AI champions, sharing prompts and wins, and turning experiments into standard ways of working that outlast any individual. For career-focused perspectives your team might value, see how to position yourself as an AI-ready leader.
Frequently asked questions
Should we make AI champions sign longer notice periods?
That rarely solves the underlying problem. Spreading skills and documenting workflows protects you far better than contractual restrictions.
How do we document AI workflows without it becoming a burden?
Keep it light: a shared document per workflow with the steps, the prompts and the review points, updated when something changes.
Is it worth training people who might leave?
Yes. Untrained staff are a bigger risk than trained staff who might leave, and development is one of the strongest reasons people stay.