You can buy a good AI tool, set it up well, and still get nothing from it, because the people who were meant to use it quietly carry on as before. Advisers who have watched many businesses adopt AI say the same thing: the hard part is rarely the technology. It is people, habits and trust.
So how do you bring your team with you? This guide covers what to say, who to involve, and how to build skills without sending everyone on a course.
Key takeaways
- Talk about what AI means for people's work before you talk about savings.
- Involve the people who do the task in choosing and shaping the tool.
- Build skills in real work: short challenges, show-and-tell, shared tips.
- Make it normal to talk about AI mistakes as well as wins.
- Give people a safe, approved way to use AI so they do not go around you.
Why do people resist AI tools?
Mostly for sensible reasons. They worry about their jobs. They have seen AI get things wrong and do not want to be blamed for its mistakes. They are busy, and learning a new tool feels like extra work on top of the old job. And sometimes they can see problems with the tool that nobody asked them about.
Resistance is rarely loud. It looks like people saying the tool is great and then not using it, or using it and redoing all its work by hand. Either way, you pay for the tool and get none of the benefit.
What should you say, and when?
Have the people conversation first, and the money conversation second. If the only thing staff hear about AI is "efficiency", they will hear "am I next?". Silence is read the same way.
Be specific and honest:
- What the tool will do: "It will draft replies to delivery queries."
- What stays with people: "You decide what gets sent, and you handle anything unusual."
- What changes in their role: which tasks shrink, which grow, and how their work will be judged.
- What you do not know yet: say so. People trust "we are not sure yet, and we will tell you when we are" far more than false reassurance.
Keep talking as it rolls out, in short regular updates rather than one big announcement.
Who should be involved?
The people who do the work, from the start. They know where the time really goes, they know the awkward exceptions, and they will spot why a tool will not work in practice long before a supplier does.
Practical ways to do this in a small business:
- Ask the team which tasks they would most like help with, and start there.
- Use their real examples to test tools, not the supplier's demo.
- Give one respected person the role of champion: not necessarily the most senior, but someone others go to for help. Give them a little time each week for it.
- Include a sceptic in any trial. If you can win them round, the rest will follow; if you cannot, they will tell you why.
How do you build skills without a training programme?
For most small firms, the best training happens inside real work. Formal courses are easily forgotten if they are not tied to the job. Try these instead:
| Approach | What it looks like | Why it works |
|---|---|---|
| Short challenges | "This month, can we halve the time spent on the weekly report?" | A clear goal on real work, with a result people can see |
| Show-and-tell | Ten minutes in the weekly meeting for someone to show what worked | People learn fastest from colleagues |
| Find the error | The team picks apart a piece of AI output on a real task | Shows where AI goes wrong and why checking matters |
| Shared tips | A shared document of prompts and lessons that worked | Stops everyone learning the same thing separately |
| Protected time | A regular hour to experiment on a low-risk task | Learning does not get squeezed out by the day job |
Aim for confidence, not expertise. Everyone should understand what the tools are good at, where they go wrong, and what information must never be put into them. Very few people need to understand how they work inside.
An example from a small firm
One small digital marketing firm needed its team to learn AI tools without slowing down client work. Rather than running formal training, it set small, time-boxed challenges, such as speeding up a regular piece of work or cutting down manual reporting, and held a weekly show-and-tell where people shared what had worked. Delivery carried on as normal, and over time using the tools simply became part of how the team worked.
A larger construction software company took a similar line when AI started reshaping roles. Its first instinct was to hire new people. Instead it invested in its existing team first, with weekly clinics and a regular innovation hour open to everyone. Sharing what they learned across teams became what drove adoption.
What about people who are already using AI on the quiet?
They almost certainly are. Staff often start using free AI tools on their own before any official decision is made. Banning this tends to drive it underground, where you cannot see what information is being pasted where.
A better response is to treat it as useful information: it shows you where people feel the need. Then give them a safe, approved route that is easier than the unofficial one, with clear guidance on what can and cannot be shared. Our guide to keeping a human in charge has more on simple rules for AI tools.
A checklist for your next AI roll-out
- Explain why you are using AI, in terms of people's work, before talking about savings.
- Ask the team which tasks they want help with.
- Choose a champion and give them some time each week.
- Test on real work, with at least one sceptic involved.
- Write down the new way of doing the task, and which old steps stop.
- Set aside time for learning, not just for using.
- Share wins and mistakes openly.
- Check after a few weeks whether people are actually using it, and ask why if not.
Watch for hidden extra work
One last thing. AI often creates new work as well as removing old work: drafts to check, more options to choose between, more output to review. If you do not plan for that, the time you hoped to save disappears into checking, and people end up more stretched than before. Ask the team how the new way of working feels, not just whether the numbers have moved. Our guide to telling whether AI is paying off explains how to count that review time.