Upskill or Be Out Skilled: Staying Relevant in the Age of AI

Is artificial intelligence replacing jobs or changing them?

Artificial intelligence is changing tasks inside many jobs, while the effect on employment depends on how organisations adopt it and which human responsibilities remain.

The International Labour Organization’s 2025 update on generative AI and jobs uses task level analysis rather than a list of jobs that will disappear. Exposure is not the same as replacement. A task may be automated, assisted or redesigned while accountability remains with a person.

That distinction makes the useful question more specific. Which parts of your work can a tool support, which parts need human judgement, and which new controls will the employer expect? A claims analyst, actuary, recruiter or data engineer can all use the question without pretending that one forecast applies to every occupation.

Why does upskilling matter in 2026?

Upskilling matters because employers need people who can combine changing tools with domain knowledge, judgement and communication.

The World Economic Forum Future of Jobs Report 2025 sets out its training expectation as a breakdown rather than a single number. For a representative sample of 100 workers, employers expect that 41 will not require significant training by 2030. Of the rest, 11 will need training that is not expected to be accessible to them, 29 will be upskilled within their current roles, and 19 will be reskilled and redeployed elsewhere in the organisation. The 11 is the figure worth sitting with: it describes people whose employers already expect the development they need will not reach them.

The same report names analytical thinking as the leading core skill, followed by resilience, flexibility and agility, and by leadership and social influence. Technology skills grow fastest, but they grow alongside those, not instead of them. Our guide to how the hiring landscape has evolved covers what that means for employers.

For a specialist professional, relevance is built at the intersection of three capabilities. You need enough technical literacy to use or challenge a tool, enough domain knowledge to recognise an implausible result, and enough communication skill to explain the implication to a decision maker. Remove any one of the three and you create risk.

What skills should professionals develop first?

Start with the skills closest to the decisions you already influence, then add the technical knowledge needed to improve those decisions.

An actuary might learn how to validate an automated data pipeline or explain a model limitation. A recruiter might learn responsible use of search tools, evidence scoring and privacy controls. A data analyst might improve experiment design, stakeholder writing and documentation. The best first skill is not the trendiest one. It is the gap that would make your current work more reliable or more valuable.

Write a short skills map with four columns: current task, tool that could assist, judgement that must remain human, and evidence of competent use. This prevents a course list from becoming the plan, and gives a manager a clear way to support development.

How can you learn artificial intelligence without becoming a specialist engineer?

Learn enough to use, test and explain the tools relevant to your role, without claiming expertise you do not have.

Begin with safe experiments. Use a public or synthetic dataset, compare a manual process with an assisted process and record where the tool was wrong or uncertain. Practise checking outputs against source material. Learn basic concepts such as training data, prompts, evaluation, access control and data leakage.

Then apply the learning to a real workflow under supervision. A useful example is a reporting task where the tool drafts a summary while you verify calculations, identify missing context and approve the final wording. Keep a short record of the checks. That record is stronger evidence than a certificate with no application.

How should employers support responsible upskilling?

Employers should provide protected learning time, safe environments, clear use cases and a review process for quality, privacy and bias.

The OECD report on bridging the AI skills gap distinguishes advanced AI skills from general AI literacy. Most teams need both. A small group may build systems, while a much larger group needs to understand how to use, question and escalate them.

Set a learning objective connected to a business decision. Define what a person may test, which data may be used, who reviews the result and how success will be measured. Do not make training a substitute for governance. A tool that saves time but creates unreviewed errors is not an improvement.

Six practical steps to stay relevant

Use a small, repeatable cycle of audit, learn, practise, verify, explain and review.

1. Audit your tasks. List recurring tasks and mark those that are repetitive, judgement heavy, sensitive or dependent on external data. This shows where automation might help and where controls are essential.

2. Choose one meaningful gap. Select one skill that improves a current responsibility. Avoid collecting unrelated badges. A focused gap is easier to practise and easier for an employer to assess.

3. Use a safe practice case. Work with synthetic or public information. Keep client, candidate and employer data out of unapproved tools.

4. Compare the output. Check accuracy, completeness, bias, explainability and time saved. Record what the tool did not understand.

5. Explain the result. Write a short note for a non technical reader stating the recommendation, evidence, uncertainty and next control.

6. Ask for feedback. Invite a manager or experienced colleague to challenge the method. Revise the workflow before presenting it as a capability.

How should professionals show AI related capability to employers?

Show a specific improvement, the checks you performed and the decision that became better informed.

Explain the starting problem, the tool or method used, the human review and the result. If the experiment failed, say why and what you changed. Employers should value responsible judgement rather than enthusiasm for every new product.

In our specialist searches, strong candidates connect technical change to role context. They explain how an AI assisted process affects underwriting, claims, actuarial work, risk or recruitment, and where a qualified person must remain accountable. Our data analyst recruiters show how that evidence is assessed in practice, and our cloud and data engineering recruitment practice covers the platform side of the same question.

What should employers avoid?

Employers should avoid vague promises, compulsory tool adoption without training, and performance measures that reward speed while ignoring quality.

Do not assume that a person who uses a consumer tool understands model risk, privacy or bias. Do not announce that a job is safe or doomed without examining its tasks. Do not ask employees to place confidential information in a system before the security and retention rules are clear.

Our guide to the impact of AI on recruitment covers governance when technology is used in hiring. This article focuses on the capability people need as their own work changes.

What can a practical ninety day upskilling plan include?

Use the first month to understand the tool and its risks, the second to apply it to a supervised workflow, and the third to demonstrate a measured improvement.

In the first month, define the task, the approved data and the quality checks. In the second, compare assisted and manual outputs on a small sample and ask a colleague to challenge the result. In the third, document the time saved, errors found, judgement retained and next control. If there is no measurable benefit, stop or redesign the experiment. A disciplined decision not to deploy is evidence of capability too.