AI reskilling moves people into new roles, while upskilling deepens current ones. Learn how to map roles, train and measure results.
AI reskilling means moving someone into a different role, while upskilling deepens their skills in the role they already hold. That distinction opens a how-to guide from Careerminds, an outplacement firm, titled AI reskilling: How to prepare your workforce, published around 17 August 2026.
The guide suggests mapping each role by how much of it AI can handle and which new skills the remainder needs. It then recommends measuring performance in the new role, retention and cost against external hiring.
This article explains that approach in plain steps, with our own examples, which are marked as such. It is based on a search extract, so it keeps to the points that extract stated. Careerminds helps companies manage layoffs and career transitions, so its advice has a commercial angle, which we discuss at the end. The guide is aimed at employers, but workers can use the ideas to plan their own moves.
Upskilling and AI reskilling: the difference that matters
The guide begins with a clear definition. Upskilling deepens skills in a current role. AI reskilling moves someone into a different one.
The difference matters because the actions, costs and promises are different. Upskilling is about making people better at what they already do. For example, a marketing analyst learns to use AI tools to speed up reporting. AI reskilling is about changing what people do. For example, a data-entry clerk whose tasks are largely automated trains to become a data quality reviewer.
Mixing the two causes confusion. If you tell employees that they will be upskilled when you really plan to reskill them into other roles, trust suffers. If you reskill when a little upskilling would do, you waste money and disrupt lives.
Workhuman, in this catalogue, advises pairing the two. Paylocity gives guidance on upskilling for AI. Careerminds adds a method for deciding which roles need which treatment.
Think of it as a decision at the level of each role. Some roles will change at the edges and need upskilling. Others will change so much that people need new roles, and AI reskilling is the answer. The next section explains how to tell the difference.
Map each role by how much AI can handle
The guide's key method is to map each role by how much AI can handle. This is a practical way to turn vague worries into specific plans.
For each role in your organisation, list the main tasks. Then, for each task, judge how much of it AI could handle now or in the near future. Some tasks, such as sorting data, drafting routine messages or summarising documents, may be largely automatable. Others, such as negotiating with a difficult client, comforting a distressed patient or making a judgement call with incomplete information, are less so.
Then estimate the share of the role that AI could take on. A role where AI can handle a small share may need only upskilling. A role where AI can handle a large share may need AI reskilling, because the human part that remains is a different job.
Be careful with the judgement. AI capabilities change fast, and estimates are uncertain. Involve people who do the work, since they know the details. Use ranges, not exact percentages. Review regularly.
Our own illustration: a customer support role might be 40% routine queries that AI can answer, 40% complex cases that need human judgement and 20% relationship tasks. The role may shift toward complex cases and relationships, with fewer people needed for the routine part.
The map gives you a shared starting point for planning.
Identify the new skills the remainder needs
The second step in the guide is to identify which new skills the remainder of the role needs. Once you know what AI can handle, you must ask what is left for humans and what skills that requires.
The remainder is often more demanding in human terms. It may involve judgement, empathy, creativity, problem solving, relationship building and oversight of AI tools. It may also need new technical skills, such as checking AI results, managing data quality or configuring tools.
Our own illustration: if AI handles the routine queries in a support role, the human remainder might need skills in handling emotionally charged conversations, solving unusual problems, spotting patterns in complaints and giving feedback to improve the AI tool.
For each role, write a short list of the skills needed. Compare it with the skills of the current staff. The gaps become your training agenda. This links to Workhuman's skills-gap audit, covered elsewhere in this catalogue.
Be specific. A skill such as communication is too vague. Better to say explaining technical issues to non-technical customers, or de-escalating angry callers.
Involve employees. Ask them what they think the future role will need, and what they would like to learn. Their ideas can improve the plan, and their involvement builds buy-in.
The result is a picture of what the future role looks like, and what people need to get there.
Choose between upskilling, AI reskilling, hiring and partnering
With the role map and skills list in hand, you can choose actions. For each role or group of roles, there are several options.
Upskill where the role stays broadly the same, but the tools change. Training should be practical and close to the job.
Use AI reskilling where the role changes substantially or shrinks, and where people could move into new roles that use their strengths. This needs more investment, such as longer training, mentoring and a transition period.
Hire where the skills are too specialised to build in time, or where growth needs new capacity. The guide suggests comparing the cost of internal moves with external hiring.
Partner where it makes sense. Training providers, universities and consultants can supply expertise, especially for complex skills.
Do nothing, or manage a reduction, where the work disappears and no suitable role exists. This is the hardest option. If it applies, handle it fairly, with notice, support and, where appropriate, help to find new jobs. Legal duties may apply, such as notice rules in some places. In the US, the WARN Act requires 60 days' notice for certain layoffs by larger employers.
Combine options. Many organisations will use a mix.
The choice should be based on evidence, not on assumption. The measures in the next sections help.
Measure performance in the new role
The guide advises measuring performance in the new role. This is how you learn whether AI reskilling works.
Define what good looks like. For each new role, set clear goals and indicators, such as quality, speed, customer satisfaction and error rates. Make sure that they are realistic for a person who is still learning.
Compare with a baseline. How do reskilled employees perform compared with people hired externally, or with their own past performance? Allow time for a learning curve, since early results may be weaker.
Collect feedback. Ask managers, colleagues and the employees themselves how it is going. Learn what supports are working, and what is missing.
Be fair. Do not judge people harshly for early mistakes. They are learning a new job. Provide coaching and adjust.
Track over time. Check performance at three, six and twelve months. Look for improvement.
Use the data to improve the programme. If many people struggle with a particular skill, add training. If a role is too hard, consider changing it.
Share results. Show employees and leaders what the data shows. Success stories build confidence and encourage others.
Remember that numbers do not tell the whole story. Listen to people's experience as well.
Measure retention and cost against external hiring
The guide's last measures are retention and cost against external hiring. They help you see whether AI reskilling is worth it.
Retention asks whether reskilled people stay. If they leave soon after training, the investment is lost. Higher retention among reskilled staff is a sign of success. It may reflect gratitude, a sense of growth and trust in the employer.
Cost compares the full cost of AI reskilling with the cost of hiring someone externally. External hiring costs include recruitment fees, advertising, interview time, onboarding, ramp-up and the risk of a poor fit. AI reskilling costs include training, time away from work, coaching, and lower productivity during the transition.
Our own illustration: if reskilling a customer support agent into a data quality role costs the equivalent of three months' salary in training and reduced productivity, and hiring externally would cost four months' salary in fees and ramp-up, the internal route saves money and keeps knowledge.
Include less tangible factors. Internal moves keep company knowledge and culture. External hires bring fresh ideas.
Be honest about uncertainty. Costs vary, and savings are not guaranteed. Track real numbers.
Compare results across groups to learn what works best, and refine the approach.
Communicating about AI reskilling with honesty
How you talk about AI reskilling is as important as what you do. People will have questions and fears.
Be clear and honest. Explain why the organisation is changing, what roles may be affected and what support will be offered. Avoid vague reassurance. Gallup finds that fear of job loss is higher where AI is in use, so silence will not calm people.
Involve managers. They are the main source of information for employees. Brief them well, and equip them to answer questions.
Listen. Run surveys, hold forums and open channels for questions. Paylocity's advice to survey employees about concerns applies here too.
Be fair in selection. Explain how people are chosen for reskilling. Use clear criteria, such as skills, interest and potential. Avoid favouritism.
Offer choice where possible. People are more committed when they have a say.
Support wellbeing. Change is stressful. Offer counselling, flexible hours during training and recognition of effort.
Follow through. If you promise training, deliver it. If you promise roles, make sure they exist.
Honest communication builds trust, which is the foundation for successful change. Without trust, even the best programme may fail.
What workers can do to prepare for AI reskilling
Workers need not wait to be told. You can look at your own role through the same lens.
List your main tasks. Which could AI handle now or soon? Which need human judgement, relationships or creativity?
Identify the skills that will matter more. Strengthen them, through practice and learning.
Explore adjacent roles. Which roles in your organisation use similar skills but are less exposed to automation? Talk to people in those roles and ask about their work.
Talk to your manager. Share your interest in learning and moving. Ask about training and opportunities.
Build evidence. Keep records of your results and new skills. They support your case for a move.
Learn about AI tools. Use them in your work where allowed. Understand their strengths and weaknesses.
Expand your network. Contacts inside and outside your organisation can alert you to opportunities.
Plan finances. Change can bring uncertainty. A small cushion helps.
If your employer offers AI reskilling, take part with an open mind. It may be an opportunity to move into a stronger position.
If not, consider whether to plan your own path, in a market where hiring is slow and competition is high.
Common mistakes in AI reskilling programmes
Reskilling programmes can go wrong in predictable ways.
No clear link to business needs. Training that does not lead to real roles frustrates people.
Poor selection. Choosing people by convenience, not by fit, leads to failure.
Too little support. Short courses without coaching or practice rarely work.
Ignoring managers. If managers are not on board, they may block moves or fail to support learners.
Unrealistic promises. Telling people that all jobs are safe, when some are not, destroys trust.
No measurement. Without data, you cannot tell what works.
Over-reliance on technology. Learning platforms help, but do not replace human guidance.
Neglecting fairness and law. Selection and layoffs must follow legal rules, including those on discrimination. Laws on AI in employment decisions in some US states add further duties.
Starting too late. Waiting until roles disappear leaves little time.
Treating it as a one-off. Skills needs keep changing. Build ongoing programmes.
Use this list as a checklist when designing or reviewing a programme. A little attention to these points can save much wasted effort.
How to read Careerminds' guide with care
Careerminds is an outplacement firm. It helps organisations manage layoffs and helps people move to new jobs. Its guide on AI reskilling is educational, and it also positions the firm as an expert in transitions. The advice is sensible, but compare it with other sources.
The extract is short and gives only the main steps. It does not offer data on results. It is a how-to guide, not research.
The date, around 17 August 2026, is approximate.
The method of mapping roles by AI exposure is helpful but uncertain. Judgements about what AI can do change quickly, and mistakes are possible in both directions. Update the map often.
The guide is likely to be aimed at US employers. South African organisations should consider local law, including labour law on retrenchment and consultation, and skills development rules.
Finally, avoid treating reskilling as a cure for all job losses. Some people will not be able to move into new roles, and some roles will not be replaced. Plan for fair treatment in those cases too.
Conclusion
AI reskilling moves people into different roles, while upskilling deepens skills in the current one. Careerminds suggests mapping each role by how much AI can handle, identifying the new skills the remainder needs, and then measuring performance in the new role, retention and cost against external hiring.
Add honest communication, fair selection and ongoing review, and you have a strong base. Workers can use the same method on their own roles.
Read the original guide for more detail, and adapt it to your own organisation.
Key takeaways
- Upskilling deepens skills in a current role, while AI reskilling moves someone into a different one.
- Map each role by how much AI can handle and which new skills the remainder needs.
- Measure performance in the new role, retention and cost against external hiring.
- Communicate honestly, select fairly and review often.
Study flashcards
- What is the difference between upskilling and reskilling?
Upskilling deepens skills in a current role, reskilling moves someone to a different role - How does Careerminds suggest mapping roles?
By how much AI can handle and which new skills the remainder needs - What should you measure after reskilling?
Performance in the new role, retention and cost against external hiring - Why should you communicate honestly about reskilling?
Honest communication builds the trust that change depends on
Summary
