Why AI rollouts fail, and how to make adoption stick

AI adoption asks people to change how they work. They may be curious, cautious, stretched for time or unsure what it means for their role. Understanding those experiences, and giving people a say in what happens next, is where useful adoption begins.

This is written for leaders who have already spent money on AI tools and cannot point to what changed, and for leaders who are being careful and want to get the foundations right before they start.

Why licences and a training day are not evidence

A licence tells you someone can open a tool. A login tells you they opened it. Neither tells you whether a report took less time to write, whether the quality of the writing held up, or whether the checking afterwards ate the time that was saved. A good training day shows people what a tool can do. It does not tell them which parts of their own week are worth changing.

That gap can leave people unsure where to start or which uses are allowed. A usage dashboard may show activity without showing whether the work improved. Discussing roles, expectations and support makes those questions visible before deciding what to do next.

Understand the person, then explore their work

Start by listening to the person doing the work. What matters to them? Where does their day become difficult? What would they like help with, and what concerns them about AI? Then explore the opportunities across their role together. The AI Adoption Programme brings our experience in leading people and change to that work over three months.

Each person's working manual records how they work best, in their own words and under their own sharing choices. Use what they choose to share to shape learning, practice and support, then ask whether that support is helping. Managers review agreed actions and outcomes; private manual answers stay under each person's sharing choices.

Agree the rules of the game

Unclear permissions can hold back useful experimentation. Leaders are accountable for the rules, and people should help make them usable: which tools and accounts are approved, what kinds of information may be used with them, how outputs get checked, and who is responsible for what. Written down where everyone can see them, the rules replace a hundred private guesses. There is more detail in how to agree AI rules at work.

Be tool-agnostic on purpose

Assess tools against the work, the organisation’s requirements and the support people need to use them. We can build around preferred or approved tools, or help you assess suitable options. Review the results in real work before deciding what to keep.

Turn it into a rhythm, then measure honestly

Habits form through repetition, not enthusiasm. Agree who owns what, put support and review in the diary, and drop the things that add nothing. Dropping them counts as progress, because the goal is better work against the priorities that matter, not more AI use for its own sake.

• Take a baseline before you change anything, so you have something to compare against.
• Measure useful adoption, meaning the work that actually changed, not logins or attendance.
• Measure time, including the review and checking effort, not just the drafting time.
• Measure quality, because faster work that has to be redone is not a saving.
• Be honest about the difference between freed capacity and cash saved.

Where to start

If you are already paying for AI tools without much to show for it, or you want solid foundations before you introduce anything, the AI Adoption Programme sets out how the three months work. A workshop can be a useful part of that, but a single day is not the same as sustained adoption.

Common questions