Workforce planning needs to see the work inside the role. A person who once assembled a report may now have to challenge an AI-generated explanation and decide whether its evidence is sound. The organisation owes that person a clear standard and time to learn.
Begin with the responsibility that is changing
The World Economic Forum estimates that 59 out of every 100 workers will need training by 2030. The practical response begins with the responsibilities people will inherit. Reviewing an AI recommendation requires domain knowledge, access to the supporting evidence and permission to reject it.
Study that responsibility with the person doing the work. Ask which errors would be difficult to recognise and where they would seek help. A training plan built around those answers will be more useful than a general introduction to a new tool.
Workforce planning can then show where review capacity needs to grow as routine preparation changes. Give managers a way to develop that capacity before altering the staffing plan. Employees should be able to see what good performance will look like in the revised role.
Describe the work before scoring the person
Start with a real workflow. Identify the tasks involved and the decisions that require judgement. Ask the people doing the work where they compensate for missing information or unclear responsibilities. Those informal adjustments often explain why a process works despite its documented design.
Then describe the capabilities the revised workflow requires. For example, give a service adviser responsibility for assessing an exception that previously arrived as a routine response to assemble. Define the controller’s responsibility for challenging a prepared variance explanation, including the evidence needed to accept it. The change should be expressed in observable work, not a vague requirement to become more AI literate.
A skills profile can support this discussion when people can correct it. Give employees a way to add evidence and explain experience the system has missed. Managers then have a richer account of capability to use in development conversations.
In the EU, Article 4 of the AI Act has applied since 2 February 2025; its July 2026 revision requires providers and deployers to support staff AI literacy in the context of their work. The UAE’s April 2026 agentic-AI framework likewise commits to continuous specialised training for federal employees.
Build learning around supervised practice
A course can introduce a capability, but performance develops through use and feedback. Give people representative tasks with access to someone who can assess the result. Include cases where the correct response is to question the assistant or stop the process.
Define what competence looks like before asking a manager to sign it off. For a review role, require the person to trace a recommendation to its sources and recognise when an exception requires escalation. The standard should match the consequence of the work.
Allow time for that practice. If training is added to a full workload without adjustment, completion figures may look healthy while capability changes little. Managers need a realistic plan for coverage and a way to discuss difficulties without making people feel that honest questions count against them.
Connect workforce choices to the operating plan
Picture a controller’s first month in the revised role. They practise on past exceptions with an experienced colleague, learn to challenge a plausible explanation and gradually take responsibility for live reviews. Their manager can see which judgements they handle confidently and where another supervised session would help.
The workforce plan records that progression. It shows capabilities developing through work, connecting the next operating requirement to people who want and are ready to take it on. A temporary specialist can support the team while internal capability grows, with the transfer of knowledge built into the assignment.
That model should support managerial judgement rather than replace it. Availability, employee interests and local operating conditions matter alongside the skills recorded. A match on paper does not establish that a move is practical or desirable.
Review the plan against actual outcomes. Look at the quality of work and the burden placed on reviewers, not only course completion or tool usage. Where the revised process creates repeated uncertainty, adjust the process or support model instead of assuming that another training session will solve it.
Make room for supervised practice
Select one role affected by an AI service this quarter and map its revised work with the people doing it. Agree the competence standard and protected practice time before changing the staffing plan.
A NectarGlobal transformation diagnostic can help your managers connect that learning plan to the operating changes their teams are being asked to make.