Giving people access to AI takes a day. Changing how they work with it takes practice.

I often hear this comment and challenge when I speak with leaders in my AI Practice Hub: We have more AI tools and faster output, but the work has not become easier. On paper, the rollout looks like a success. Licences are in place, training sessions have been delivered and usage numbers are going up.

When I ask people how their week feels, the answers sound different. People tell me they spend more time checking AI output than they used to spend writing. Managers receive documents which they consider to be made with AI, and they don’t trust this work anymore. Some employees have stopped using the tools after one bad experience. Others trust the output too much and pass it on without reading it properly.

This is the gap is real: AI has made individual tasks faster, but the way the work is organised has stayed the same. When that happens, speed often turns into extra volume, extra checking and extra pressure.

 

 

Why this happens

 

When I look closer at these situations, I usually find three causes.

AI is added on top of the old workflow. Teams receive a tool, but nobody redesigns how the work gets done. The old steps stay in place, and AI becomes one more step. People draft with AI, rewrite everything in the old format, then send it through the same review loop as before. The effort moves around, but it does not go down.

Responsibilities are unclear. When nobody has defined what AI should do and what people should do, employees decide this on their own. Some become overly cautious and check every sentence. Others assume the tool is right because it sounds confident. Both lead to more effort and less confidence in the result.

People do not feel safe enough to be honest. Many organisations encourage experimentation in their AI policy. In daily work, however, employees are measured on speed and accuracy. If someone shares a report with an AI error in it, the question is often “how did you miss this?” rather than “what can we learn from this?” Under that pressure, people either avoid the tool or keep their problems to themselves. Neither helps the organisation learn.

All three causes sit in how work is organised and led, which is why a better tool will not solve them. When I help leadership teams move from AI access to AI adoption, I work with three building blocks. They build on each other, and all three need attention at the same time.

 

 

 

The approach: Workflow, Trust, Practice

 

1. Workflow: decide who does what

Adoption starts with the workflow. Before asking whether people use AI, ask where in the process AI should help. A simple rule works well in most teams. Let AI bring information together, find patterns and prepare a first draft. Let people add context, check what looks wrong and make the decisions.

Take a weekly report. AI can pull information from different sources, compare it with last week and highlight what has changed. That saves time. But the person writing the report still needs to understand why those changes happened, whether the data behind them is reliable and what the team should do next. Those three questions are where the value of the report lies, and they remain a human responsibility. When this split is clear, people know where to rely on AI, where to question it and which part of the work is theirs. Checking becomes focused on the parts that matter.

 

2. Trust: make it safe to try and to speak up

Adoption also depends on trust. Three questions tell you a lot about where your team stands:

• Can employees try the tool on a small task without pressure to get it right the first time?
• Can they say openly when it does not work?
• Do they have a voice in how it is used in their team?

If the answer is no, a policy encouraging experimentation will not help much. People will do what feels safe, and that usually means avoiding the tool or hiding the problems they run into. Leaders shape this more than any guideline. When a manager says “I tried this, it got the numbers wrong, and here is what I changed”, the team learns that mistakes are part of the process. When a manager only asks who is to blame, the team learns to stay silent.

 

3. Practice: learn in small cycles

A rollout can happen in a day. Changing how people work takes practice: try, learn, adjust, then build on what works. This is where many organisations lose momentum. They treat adoption as a project with an end date, when in fact it is a habit that grows over weeks and months. Teams need small, repeated moments to test something, see what happens and improve it.

Practice also needs time. If AI is expected on top of a full schedule, people will fall back on what they know. Adoption grows when a new way of working replaces an old step, so people feel the time saving from the start.

 

 

 

Practical recommendations

 

Here is how to put each building block into action.

 

For the workflow

• Choose one recurring task your team does every week, such as a status report, a customer summary or a meeting preparation.
• Write down the steps as they happen today, then mark which ones AI can take over: collecting information, comparing data, preparing a first version.
• Mark the steps that stay with people: explaining causes, judging data quality, recommending actions and making decisions.
• Define two or three checks the person should always do before sharing the result. For example: does the data source make sense, do the numbers match what I expect, can I explain the conclusion in my own words?
• Remove the old step once the new one works, so the time saving becomes visible.

 

For trust

• Start with low-risk tasks where a mistake has no serious consequences, so people can learn without fear.
• Add a short question to your weekly team meeting: where did AI help this week, and where did it not?
• Share your own experiences as a leader, including the ones that did not go well.
• Involve the team in deciding how AI is used in their workflow. The people doing the work usually know best where it helps.
• Make it clear that the person who shares a result owns it, and that raising a doubt about AI output is welcome.

 

For practice

• Work in two-week cycles: try one change, review it with the team, then adjust.
• Keep a short shared list of what works, including the prompts that gave good results.
• Ask one or two colleagues who feel confident with AI to support others in the team.
• Build on success. Once one workflow works well, move to the next one, rather than starting ten at once.

 

 

 

What corporate leaders can do now

 

• Pick one workflow per team and agree on what AI does and what people do in it.
• Name a human owner for every AI-supported result, so responsibility is always clear.
• Define simple checks people run before sharing AI output.
• Create room to try on small, low-risk tasks, and protect time for it in the schedule.
• Ask openly what did not work, and treat the answer as useful information.
• Lead by example by sharing your own use of AI, including your mistakes.
• Review every two weeks and build on what works before scaling to more teams.
• Measure whether the work has become easier, alongside how much faster the output is.

 

The question for you is simple: what small change would help your team use AI well this week?