How Do I Get My Team to Actually Use the AI Tools We've Bought?

21 May 2026

Why Your Team Is Not Using the AI Tools

The blunt answer is that most teams do not ignore AI because they are lazy or resistant to progress. They ignore it because the business bought a tool and assumed adoption would happen by itself.

That almost never works. A licence does not change habits. A demo does not change workflow. A one-hour training session does not make people confident enough to use AI in front of clients, colleagues, or managers.

The UK data backs this up. The Department for Science, Innovation and Technology found in its 2026 AI Adoption Research that only 16% of UK businesses were currently using at least one AI technology, while 80% were neither using AI nor planning to adopt it. Among the businesses that had adopted AI, 77% said less than half of their staff were currently using it. On average, only 30% of staff in adopting businesses were using AI. Source: DSIT AI Adoption Research.

That is the adoption gap in plain numbers. Even when a business has crossed the first hurdle and bought or deployed AI, most staff are still not using it day to day.

The most common reason is not technical failure. It is that nobody has translated AI from a general capability into a specific way of working. People are told, "Use Copilot" or "Try ChatGPT", but they are not told which task should change on Monday morning, what good output looks like, what they are allowed to paste into the tool, or how their manager will judge the result.

Start With Workflows, Not Features

The first practical step is to stop talking about AI features and start naming the work you want to change. People do not adopt software in the abstract. They adopt a better way to complete a task they already recognise.

For most UK SMEs, the best starting point is 3 to 5 workflows where AI can remove visible friction within 30 days. Good candidates include drafting first responses to customer enquiries, summarising long email threads, preparing meeting notes, turning sales call notes into CRM updates, reviewing policy drafts, generating first-pass marketing outlines, analysing support tickets, or creating finance commentary from spreadsheet data.

Bad starting points include vague ambitions such as "be more innovative", "use AI across the business", or "make everyone 20% more productive". Those statements are too broad to change behaviour.

Write the workflow down in a simple before-and-after format. For example: before AI, account managers spend 20 minutes writing a follow-up email after every discovery call. After AI, they paste approved call notes into a secure prompt, generate a first draft, check it against the client's actual requirements, and send a polished version in 7 minutes. That gives the team a real job to practise, not a vague instruction to experiment.

Do this for each pilot workflow. Name the owner, the tool, the acceptable data, the human review step, and the success measure. If you cannot explain the workflow on one page, it is not ready for rollout.

Deal With the Real Reasons People Resist

Resistance usually has a rational core. If you ignore it, the team will quietly go back to the old way of working.

The first reason is fear of looking incompetent. Many capable people are used to being good at their job. AI makes them feel like beginners again. If managers treat early mistakes as performance issues, people will stop experimenting.

The second reason is fear of job loss. Staff may hear "AI adoption" as "management wants fewer people". You need to be direct about this. If the goal is to increase capacity, improve response times, reduce admin, or free people from low-value work, say that clearly. If the business is genuinely using AI to reduce headcount, do not pretend otherwise. People can cope with difficult news better than they can cope with evasive messaging.

The third reason is distrust of output quality. This is sensible. AI can be wrong, bland, overconfident, or out of date. The answer is not blind trust. The answer is human checking. DSIT found that 84% of UK businesses using AI reported at least some human input or checking, and 67% reported significant input or checking. That should be the default model for most business use: AI drafts, summarises, suggests, and structures, but a competent human approves.

The fourth reason is data anxiety. In the UK, this includes UK GDPR, client confidentiality, employment data, financial data, and commercially sensitive information. People need a simple policy that says what can and cannot go into each tool. A vague "be careful" is not enough.

The fifth reason is lack of time. Leaders often say they want adoption but still expect the same workload, deadlines, and output during the learning period. That is not realistic. If you want adoption, give people protected practice time. Two 45-minute sessions per week for four weeks is often enough to create momentum in a small team.

Use a 30 to 90 Day Adoption Plan

A practical rollout does not need to be complicated. It does need to be structured.

The key is to measure behaviour, not enthusiasm. Training feedback forms are useful, but they do not prove adoption. You need to know whether people are using the tool in live work, whether the output is good enough, and whether the workflow is genuinely better than before.

For a paid tool such as Microsoft 365 Copilot or ChatGPT Team, also measure cost per active user. If 40 licences are being paid for and only 9 people use the tool weekly, the real cost per active user is more than four times the headline licence cost. That number gets leadership attention quickly.

Make Managers Change Their Own Behaviour

AI adoption fails when leaders delegate it to training and carry on managing work exactly as before.

If managers still ask for the same report in the same format, review the same manual spreadsheet, and reward the same old process, the team will read the signal correctly: AI is optional theatre.

Managers need to model three behaviours. First, they should use the tools visibly. That might mean summarising a meeting with AI, creating a first-pass project plan, or using AI to compare options before a team discussion. Second, they should ask better review questions. Instead of asking, "Did you use AI?", ask, "Which part of this did AI help with, what did you change, and what did you reject?" Third, they should protect quality. AI-assisted work should still be checked for accuracy, tone, confidentiality, and judgement.

This matters because staff copy what managers reward, not what managers announce. If the fastest route to praise is still doing everything manually, people will keep doing everything manually.

One useful rule is that every manager in the pilot should bring one AI-assisted work example to a weekly review meeting. Not a perfect showcase. A real example. Good, bad, or mixed. The point is to normalise learning and make the new working pattern visible.

Put Simple Governance in Place Early

Governance does not need to be a 40-page policy. For most SMEs, a two-page AI usage standard is a better start.

It should cover five things. Which tools are approved. Which data is allowed. Which data is banned. Which outputs require human checking. Who to ask when unsure.

For example, client names, public website copy, anonymised call notes, and internal process notes might be allowed in an approved business AI tool. Payroll data, health information, legal advice, passwords, unpublished financial results, and confidential client documents might be banned unless a specific data protection review has approved the use case.

UK businesses should also be clear that AI does not remove their responsibilities under UK GDPR, confidentiality agreements, sector rules, or employment obligations. If AI is used to help make decisions about people, customers, credit, legal matters, or safety, you need stronger controls than you need for drafting a blog outline.

The aim is not to frighten people. The aim is to remove uncertainty. When staff know the boundaries, they are more willing to use the tool inside those boundaries.

When This Does NOT Apply

Do not force AI adoption just because the licences have already been bought. That is sunk cost thinking.

This approach does not apply if the tool is genuinely the wrong fit. For example, a business may buy a general AI writing tool when the actual bottleneck is messy CRM data, slow approval processes, or an outdated finance system. In that case, more training will not solve the problem.

It also does not apply if the work is too sensitive for the tool you have chosen. If your team handles regulated legal, medical, financial, safeguarding, or employment decisions, you may need a more controlled environment, stronger audit trails, and specialist advice before expanding use.

Do not push adoption if managers are using AI as surveillance or as a threat. "Use this or be left behind" is a poor adoption strategy. It creates defensive behaviour and hidden workarounds.

Finally, do not expand licences until the pilot proves value. It is better to have 10 people using AI weekly in useful workflows than 100 people with licences they open once a month.

The Practical Answer

If your team is not using the AI tools you bought, the fix is not another inspirational demo. The fix is a disciplined adoption process.

Pick a small number of real workflows. Give people exact examples. Set safe data rules. Train with real work, not generic exercises. Make managers use the tools visibly. Measure weekly active use and business outcomes. Stop the workflows that do not produce value.

techUK reported in March 2025 that lack of expertise was the top AI adoption barrier in a survey of more than 1,000 UK IT decision-makers, at 35%, followed by high costs at 30% and uncertain ROI at 25%. Smaller businesses cited high costs, uncertain ROI, and lack of expertise as key barriers. Source: techUK and ANS research.

That is why adoption needs to be practical. Your team does not need a lecture on the future of work. They need permission, examples, rules, practice, and proof that using AI will make their actual job better.

If you want an honest view of whether your current AI tools are worth saving, start with a focused workflow review. Map what has been bought, who is using it, where the friction is, and which 3 to 5 workflows are most likely to produce value. If it makes sense to continue, build the adoption plan. If it does not, cancel the licences and fix the real bottleneck first.

Is This Right For You?

This approach is right for you if you have already bought tools such as Microsoft 365 Copilot, ChatGPT Team, Claude Team, Gemini for Workspace, Notion AI, Zapier, Make, or an industry-specific AI platform, but usage is patchy. It is also right if senior leaders keep asking why the AI budget is not producing visible productivity gains.

It is not right if you have no clear business process to improve, no manager willing to own adoption, or unresolved data protection concerns. In that case, pause the rollout and fix those foundations first. Forcing people to use AI before the business has defined safe, useful work patterns usually creates more resistance, not less.

Frequently Asked Questions

How long should it take for a team to adopt AI tools?

For a small UK team, you should expect visible adoption within 30 to 60 days if the workflows are clear and managers are involved. Full habit change usually takes closer to 90 days. If usage is still weak after 90 days, the tool, workflow, training, or management signals are probably wrong.

Should AI tool usage be mandatory?

Not across the whole business. It can be mandatory for specific approved workflows once the business has tested them, documented the process, and trained the team. Making general AI use mandatory before people understand the value and risk usually creates poor-quality work and quiet resistance.

What is a good AI adoption metric?

Weekly active use by role is the first useful metric. After that, measure time saved, rework reduced, response time improved, or quality improved in a named workflow. Do not rely on licence count or training attendance as proof of adoption.

What should we do if staff are worried AI will replace them?

Address it directly. Explain whether the goal is capacity, quality, faster turnaround, cost control, or headcount reduction. Do not hide behind vague language. If the goal is augmentation, prove it by redesigning work so people spend less time on low-value admin and more time on judgement, client service, and delivery.

Do we need an AI policy before the team starts using tools?

Yes, but it can be simple at first. You need approved tools, banned data types, human review rules, escalation routes, and examples of acceptable use. For regulated or sensitive work, you need stronger governance before rollout.

Which teams should use AI first?

Start where the work is repetitive, text-heavy, and easy to review. Marketing, administration, sales support, customer service, and internal reporting are common starting points. Avoid starting with high-risk decisions about employment, legal advice, safety, credit, or vulnerable customers unless you have specialist controls.

Should we cancel AI licences if people are not using them?

Not immediately. First run a 30-day workflow pilot with a small group and proper support. If active use and measurable value still do not appear, cancel or reduce the licences. Paying for unused AI tools is not a strategy.