You paid for AI training. Why is nobody using it?

You bought the workshop. People learned clever prompts, laughed at an unexpectedly good poem and received a certificate. A month later, the tool has fewer regular visitors than the office printer's troubleshooting page. Management asks why employees are resisting change. HR is asked to organise another workshop.

Training can help people learn a capability. Adoption asks them to change a working habit under real deadlines, permissions and incentives. Those are connected problems, but buying the first does not automatically solve the second. Before scheduling another session, find out where the habit breaks.

Ask people to show you the last task

“Why don't you use AI?” invites a defence. “Show me the last report you prepared” invites evidence. Watch the inputs, the drafting, the checking and the handover. Ask which part was repetitive, which required judgement and where another person needed to approve something.

The answer might be that the tool cannot access approved information, that staff cannot use customer data in it, or that checking the output takes longer than their usual method. It might also be unfamiliarity. Treat each possibility as a different problem with a different response.

In a hypothetical HR team, employees learn to generate interview questions but continue creating them manually. Their company requires questions linked to defined job criteria, and nobody has explained how generated questions should be reviewed. The missing ingredient is a usable review method. Another slideshow of prompt tricks would leave the blockage intact.

Choose one useful habit for each role

Build training around a task people actually do. For HR, that might be producing an initial communication draft from approved facts. For an operations team, it might be turning a process note into a checklist. For account managers, it might be preparing questions from an authorised, anonymised brief.

Pick a low enough risk to allow practice and a high enough frequency to make the practice meaningful. If the task happens once a year, the habit will have a long winter. If it happens every day but involves sensitive decisions, narrow the exercise to a preparatory step that can be checked.

Write the habit as a short sequence: choose approved inputs, request a draft, check against defined criteria, revise, and hand over through the normal process. The employee should leave knowing what to do on their next working day. A catalogue of fifty possibilities can be exciting and unusable at the same time.

Teach evaluation alongside generation

People need to distinguish a useful draft from convincing nonsense. Give them examples of correct outputs, incomplete outputs and plausible errors. Ask them to explain which facts need checking and which decisions belong to someone qualified. Practise rejecting a result as confidently as producing one.

The Generative AI at Work study by Brynjolfsson, Li and Raymond examined an assistant in a customer support setting and found that its effects differed across workers. That context-specific result does not promise the same gain in your business. It does support asking which employees, tasks and forms of support your programme actually serves.

An experienced specialist may need training in reviewing edge cases, while a newer colleague needs help understanding the task itself. A shared introductory session can be useful, followed by role-specific practice. Identical attendance does not imply identical learning needs.

Resolve permission before encouraging experimentation

Employees should not have to interpret a vague security policy while a trainer enthusiastically says, “Try it with your own documents.” Explain which tools are approved, what information can be entered, who can see the output and what must remain outside the exercise. Use synthetic examples when appropriate.

The NIST AI Risk Management Framework is a useful reference for considering risks throughout AI use. For training, my recommendation is to translate your organisation's decisions into plain instructions people can apply during the task.

HR should collaborate with the people responsible for security, data and the relevant business process. A memorable rule and an accessible contact are often more practical than sending staff to a policy nobody can locate. Rules can be proportionate without being imaginary.

Give practice a place in the working week

If the team is fully booked and the manager evaluates only existing output, using an unfamiliar tool introduces risk for the employee. Make a small, explicit space for practice and say which work can wait. Otherwise, the organisation has called something a priority while refusing to allocate any time to it.

A useful manager phrase is: “Use this week's process update as the practice task. Spend a bounded session on the new method, keep the old method available, and tell us what needed correction.” People get permission to learn and a clear stopping point.

Pair colleagues who can discuss the work rather than crown the fastest prompt writer as company oracle. Hold a short review of one useful example and one failure. Keep shared instructions brief enough to update. A large prompt library becomes another place where documents go to enjoy retirement.

Answer the question underneath the question

Some employees are worried about their jobs, about losing expertise, or about becoming responsible for machine errors they cannot control. These concerns deserve a direct conversation. If roles might change, say what is known, what remains uncertain and when the next decision will be communicated.

Do not promise universal job safety if leadership cannot support that promise. Equally, avoid turning every training exercise into an implicit redundancy test. Explain how the programme will be evaluated and discuss how any released capacity would be used. Trust needs consistent behaviour after the workshop.

When someone says the tool is useless, ask for the example. They may have discovered a limitation that protects the company from a costly mistake. Curiosity gives you better information than immediately assigning them a motivational module.

Measure changed work rather than enthusiastic attendance

Keep completion and satisfaction measures, but add evidence of task competence. Can the employee choose suitable inputs, obtain a usable draft, detect errors and follow the handover rules? Observe a small sample of real work, with appropriate privacy protections, rather than ranking people by prompt counts.

For a hypothetical trial, agree to review comparable reports before and after training. Note preparation time, corrections and the recipient's assessment. If quality stays acceptable and the full task gets easier, investigate whether the method is repeatable. If it fails, diagnose the reason before requiring everyone to adopt it.

My corporate AI training experience sits alongside product and implementation work, because a useful workshop needs contact with the work that follows. Talk to me about your team's actual tasks. We can discuss role-specific practice, manager involvement and a sensible adoption test. Start with my practical AI adoption guide if you are still choosing the first workflow.

Questions people ask

Why are employees not using AI after training?

They may lack approved access, a useful task, time to practise or a clear way to check outputs. Some may reasonably worry about confidentiality or changes to their role. Observe the last real task instead of assuming resistance. Identify whether the problem is skill, permission, workflow or incentives, then choose a response that addresses that specific barrier.

How should HR measure the success of AI training?

Measure whether staff can safely complete suitable tasks, judge outputs and follow the required review process. Attendance and satisfaction provide useful context but do not establish adoption. Review a small, appropriate sample of comparable work for preparation time, corrections and quality. Avoid using prompt counts as a performance ranking, because frequent tool use can occur without useful results.

Should every employee receive the same AI workshop?

A shared introduction can establish common boundaries and terminology, but practical training should reflect different roles and experience. Choose tasks that are frequent, relevant and safe enough to practise. Experienced staff may need help evaluating exceptions, while newer colleagues may need stronger task foundations. Agree with managers how practice fits the working week and who supports questions afterwards.