Build an AI skills plan around the people already in your business

Nov 7

Start an AI skills plan with the work your people already do. Identify a task, define a safe standard and give employees supported practice before judging whether the new workflow belongs in everyday operations.

Why training belongs in the workforce conversation

Developing people before a vacancy appears is a central argument in Daniel Orelowitz’s workplace-training commentary. He also identifies AI as one of the technologies changing existing roles. These are the author’s arguments, not a measured forecast of South African AI vacancies. [R1] They provide a reason to assess learning needs inside your business rather than waiting for a new job title to appear.

Map tasks instead of labelling people

Ask each team to list recurring outputs: a customer reply, a report, a meeting follow-up or an internal briefing. For each output, record the inputs, confidential information, likely errors and final approver. Separate tasks where AI might assist with a draft from decisions that must remain with an authorised person. Do not use enthusiasm for a tool as a proxy for competence, or uncertainty as a reason to exclude someone from learning.

Design a small practice assignment

Illustrative example, not a source finding: an administrator practises turning fictional meeting notes into a draft action list. The learner must flag missing owners, avoid inventing deadlines and explain what was excluded. A colleague checks the draft against the notes. This tests briefing, verification and judgement together, rather than rewarding the longest prompt or most polished-looking output.

Create a fair learning routine

Protect practice time, provide approved tools and make the review standard available before the exercise. Invite staff to explain unclear instructions. Offer alternatives where connectivity or access prevents participation. Candice Solomons’s digital-talent commentary argues that adaptability and other capabilities can be developed alongside technical skills. [R4] That is a useful design principle, not evidence that a particular course guarantees retention or promotion.

Assess the output and the checking

Keep the original task, the learner’s brief, the draft and the corrected version. Ask the learner to identify one unsupported statement and show how it was checked. Discuss whether the workflow reduced effort after review, not just whether a draft appeared quickly. If performance is inconsistent, identify the specific gap: task definition, tool knowledge, source checking or subject expertise. Repeat a similar exercise before expanding the scope.

Questions and answers

Does everyone need the same AI course? No. A shared foundation may help, but role-specific practice should reflect the work and its risks. Does training guarantee staff retention? No. The source argues for investment in people; it does not establish that any particular training prevents departures.