How AI is changing the skills you already use at work

Most of your skills will still matter. You will use them differently

Learning new AI tools matters. But I think the bigger change is how you use the skills you already have.

Most of those skills are not going away. You may just need them at a different step.

Think about a manager preparing a weekly project update. They still need problem solving, communication, and judgment. AI may gather the updates, compare dates, and draft the first summary. The manager now spends less time assembling information and more time deciding what matters, what looks wrong, and what the team should do next.

The manager still uses the same skills. They just spend more time making sense of the information and deciding what to do.

What you need to know

1. Most workplace skills will still matter as AI becomes part of more jobs.

2. Problem solving, creativity, and data skills are becoming more important in many roles.

3. AI can prepare an answer, but people still need to frame the problem, check the work, and make the decision.

4. You also need to understand AI well enough to use it properly and notice when it gets something wrong.

5. The best way to build these skills is to practice them in work your team already does.

What McKinsey actually found

McKinsey's HR Monitor 2026 surveyed about 1,300 HR professionals and 5,500 employees across 10 countries. When HR professionals were asked about future skills, 44% placed problem solving in their top 5. Creativity moved into second place, and data analytics and AI remained in the top 3.

What stood out to me is that this list goes well beyond coding or learning a new AI tool. HR leaders are also looking for people who can understand a problem, come up with useful ideas, make sense of evidence, and decide what to do next.

A separate study from the McKinsey Global Institute looked at thousands of skills listed in US job ads. It found that roughly 72% are used in both work AI may do and work that still needs a person.

Jobs will still change. McKinsey expects nearly every occupation to use skills differently by 2030. The encouraging part is that you may not need to start again. In many cases, you will use what you already know at a different point in the work.

Skills do not show up one at a time

We often think about one skill at a time: communication, data analysis, problem solving.

But real work mixes them together. The same skill can show up at several points in a task, and AI may change each point differently.

Take that weekly project update. The full job includes collecting notes, checking dates, spotting risks, writing a summary, and recommending what should happen next. AI may be able to collect the notes, compare the dates, and prepare a first draft.

The manager still needs to notice that one team has stopped reporting progress. They need to understand that a small delay affects a client promise. They need to decide whether to move people, reduce the scope, or accept the risk. Then they need to explain that decision clearly.

The manager still needs all the same skills. They just spend less time gathering information and more time checking it, making decisions, and explaining those decisions.

AI can give you options. You still need to choose

AI can give you an answer in seconds. But a quick answer only helps if you asked the right question.

If you ask an AI tool how to fix a delayed project, it can give you a tidy list of options. Add more people. Change the deadline. Reduce the scope. Speak to the client.

The hard part is deciding which problem you are really solving. Is the team short of time, or is the goal unclear? Is the deadline fixed, or has nobody asked? Will adding another person help, or create more coordination work?

So your role changes. You give the tool the right context, question its assumptions, and weigh the tradeoffs before making the call.

AI may speed up the work. You are still responsible for the decision.

AI can give you ideas. You still need to set the direction

AI is very good at producing options. Ask for 20 campaign ideas, 10 product names, or 5 ways to explain a difficult topic, and you will have them in seconds.

More options do not automatically give you a better idea. Someone still has to know what fits the audience, what feels tired, what the company can actually deliver, and what is worth trying.

The creative work often starts after AI gives you the options. You decide what to keep, what to combine, and what to throw away.

Sometimes you will still create from a blank page. But when AI gives you a first draft, your knowledge of the audience and your sense of what fits are what make it useful.

Data skills are also about knowing what to trust

AI can already help clean a spreadsheet, write a formula, build a chart, and point out patterns. That can save time.

But you still need to ask some basic questions. Where did the data come from? Is anything missing? Did the tool compare the right periods? Is the pattern meaningful, or is it just unusual? What can we actually decide from this?

A dashboard can look convincing and still be wrong. A clear explanation can still be based on a weak assumption.

Working well with data now means checking what went in, understanding the limits, and deciding what the result really tells you. AI can do the calculation. You still need to decide whether the answer makes sense.

AI fluency still matters

You still need to understand the AI tool you are using. McKinsey found that demand for AI fluency in US job postings rose nearly 7 times between 2023 and 2025.

AI fluency means being able to use and manage AI tools. For most professionals, that does not require building a model or becoming a software developer. It does require knowing how to give useful context, review an answer, protect sensitive information, and recognize when the tool is outside its depth.

The report also found that some specialist technical skills ranked lower than before. I would be careful with that finding. It does not mean software development no longer matters. It only shows what HR leaders placed near the top of a broad skills list. McKinsey also says the countries in the survey changed, so the comparison with the previous year is only a rough guide.

For most people, this means understanding the tool, knowing the subject well enough to check its work, and staying responsible for the result.

A simple way to look at one task

You can make this practical in about 20 minutes. Pick one task your team repeats. Do not start with an entire job description. Choose one piece of work your team knows well.

Then answer 5 questions.

1. What should you have at the end? It might be a decision, a document, a service, or something else your team needs to deliver.

2. What can AI prepare? This might include finding information, comparing records, drafting, calculating, or suggesting options.

3. What still needs a person? Write down what someone must check, decide, explain, or approve.

4. Which skill does the person need here? It could be problem solving, creativity, data analysis, communication, customer knowledge, or something else.

5. What would better performance look like? Choose something you can notice: fewer mistakes, better decisions, faster work, or useful feedback from the people affected.

For a weekly project risk update, AI might collect notes, compare dates, and draft possible risks. The manager checks what is missing, judges which risks are serious, chooses what to escalate, and explains the tradeoffs. That shows you which skills to practice: framing the problem, checking AI's work, and explaining the decision.

If you manage a team and want help applying these questions to a real situation, the AI Adoption Working Session gives you private time to choose where AI can help, discuss suitable tools, and make a three-week plan.

What managers can do now

This changes more than what goes into a training session. It also changes how managers hire, coach, and review people's work.

Training. Let people practice with real tasks. Ask them to find mistakes in an AI-generated report based on work they already understand.

Hiring. Ask candidates to review an AI answer, explain what they would check, and talk through a decision. The quality of their reasoning may tell you more than the polish of the final document.

Development. Give people more chances to interpret information, make a recommendation, and explain why. Then give them feedback on how they made the decision.

Performance. Look beyond how often someone opens an AI tool. Pay attention to the quality of the work, how carefully they reviewed it, and what happened next.

People get better at using AI when they practice on work they already understand. This article explains why supported practice matters.

A few honest things

These skills still take practice. Good judgment comes from experience, feedback, and knowing the subject well enough to notice when something is off.

AI can also make a weak decision look polished. A confident-sounding answer may encourage people to accept it too quickly. Teams need clear review habits, especially when the work affects customers, money, safety, employment, or sensitive information.

Some technical roles will change sharply, and some tasks will shrink. People in those roles deserve clear information about what is changing and proper support to adapt.

A list like this is only a starting point. Problem solving looks different in finance and customer service. Training works best when people practice with the real tasks, data, and decisions from their jobs.

My two cents

My two cents: start with the work, then identify the skills.

A list of future skills can make sense and still leave you wondering what to do on Monday. Start with one task instead. Look at what AI prepares, what you check, and where the decision sits. That will show you what the team needs to practice.

The question I would ask is simple: after AI gives you the first answer, what do you still need to notice, choose, and explain?

Whatever comes after that first answer is probably what your team needs to practice next.

Common questions

Which skills will matter most as AI becomes part of more work?

Problem solving, creativity, communication, judgment, and data skills will still matter in many roles. The exact mix depends on your work. Connect each skill to a task you do instead of treating one list as the answer for every job.

Does everyone need technical AI skills?

Most people need to know how to use AI safely and check its work. They do not need to learn how to build an AI model. They should know how to give the tool context, protect information, review the answer, and stop when they need help.

What does human judgment mean when you use AI?

It means looking at AI's answer alongside what you know about the job, the people involved, and the risks. Then you decide what happens next and take responsibility for it.

How should a company train people for these changes?

Choose one task the team knows well and let people try AI inside it. Give feedback on the questions they ask, what they check, and the final decision. Then write down what worked so they can use it next time.

Where to start

Choose one task your team repeats every week or every month. Use the 5 questions above, then pick one skill to practice the next time the work comes around.

Keep the first test small so you have time to review it properly. The goal is to notice what changed and whether the work got better.

If you are thinking about your own job rather than a team, the three bucket exercise in this article can help you sort the tasks AI may do, the tasks it can support, and the tasks that still need you.

If you need help across a whole team, this guide to AI readiness coaching explains the different kinds of support available.

Most of your skills will still matter. The useful question is how you will use them when AI handles part of the task.

Sources: McKinsey HR Monitor 2026. McKinsey Global Institute, Agents, robots, and us: Skill partnerships in the age of AI.