The standard response to AI is often “we need to reskill”. It is correct but incomplete. Reskilling becomes meaningful only when we can explain what work is changing, which activities are growing or shrinking, and which capabilities people need to perform the new mix of work well.

Activities provide that missing connection.

Every activity draws on a combination of skills. “Prepare a management report” may require data handling, business understanding, written communication, visualisation, and quality checking. “Advise a stakeholder” may require interpretation, influence, judgment, empathy, and domain knowledge. When AI changes an activity, it changes the relative importance of those skills.

This does not mean human skills are replaced by generic “AI skills”. In many cases, augmentation increases the value of distinctively human capability. If a tool drafts a first version, the person needs stronger judgment to assess whether the draft is accurate, appropriate, and useful. If search and synthesis become faster, the differentiator becomes framing the right question, deciding what matters, and acting responsibly on the answer.

A practical skills strategy should therefore distinguish among three effects.

Skills reduced in volume, not necessarily eliminated. Routine data entry, formatting, basic retrieval, and repetitive checking may consume less time. Some proficiency remains necessary for oversight, but the organisation may need fewer people spending large portions of their week on these activities.

Skills amplified by augmentation. Analytical interpretation, critical evaluation, problem framing, stakeholder communication, and decision-making become more prominent when preparation is faster. These are not vague soft skills; they are observable capabilities linked to changed activities.

Skills required to work responsibly with AI. Employees need to understand when to use a tool, how to provide useful context, how to verify outputs, when to escalate, and how to protect confidential or sensitive information. This is not a specialist curriculum for a small AI team. It is part of modern professional practice.

An activity-to-skills map makes these priorities visible. For each activity, identify the current skills, the expected AI change, and the future skill emphasis. Then look across roles. You may find that several functions need the same cluster: data interpretation, evidence evaluation, customer problem solving, or AI-assisted content review. This supports learning investments that serve mobility across the organisation rather than narrow, role-by-role programmes.

The most credible learning message is specific: “This activity is changing. These elements will be assisted. These capabilities will matter more. Here is how you can build them.” It is far stronger than telling people to become “future-ready”.

AI may accelerate the need for learning, but it also gives organisations a better opportunity to connect learning directly to work. Activities are where that connection becomes practical.

**Next article:** If the content of work changes faster than job titles, careers must be designed differently too.

‍