AI isn't just changing jobs. It's changing what organisations should reward.

Just over seven weeks ago, I explored whether AI could make organisations more alike, even as it helps them become more productive.

This week I want to look at something closer to home for anyone managing a business or leading a team. If AI is changing what creates value, are we changing how we recognise the people creating it?

Most conversations about AI and employment still focus on which jobs will disappear and which will survive. I think we need to look at this differently.

A job is a collection of activities. Some of those activities create significant value. Others are necessary but don't necessarily differentiate the person doing them. AI is changing that balance.

The problem is that many organisations may still be measuring contribution in much the same way they did before AI arrived. And I think that deserves more attention than it is getting.

Belief is cheap. The scorecard is what shapes behaviour.

We have seen this before

Technology has always changed work. We saw it with the steam engine, electricity, computers and eventually the internet. None eliminated work itself. They changed the activities people performed and the skills that employers valued.

Economist David Autor has written extensively about this. His research explains why automation often replaces particular tasks rather than entire occupations. Some tasks disappear, others become more important, and new activities emerge.

AI fits that pattern, although the speed and breadth of its development may make this transition particularly challenging.

Think about how much professional work involves finding information, writing reports, preparing presentations, summarising meetings or analysing data. These are activities that have traditionally required time, experience and effort.

Today, AI can perform many of them remarkably well. Often faster than we can, and sometimes to a higher standard.

That doesn't automatically mean the people doing those jobs become less valuable. It means we need to reconsider where their value comes from.

The reward is moving, but the measurement hasn't

For years, organisations have used visible output as an indication of contribution. How much work did someone complete? How many reports did they produce? How many presentations did they prepare? How many activities did they record?

There was a reasonable basis for doing this. Producing good work required effort and capability. If someone delivered a well-researched presentation, you could reasonably assume they had invested time in understanding the subject.

That assumption is becoming less reliable. The deck no longer proves the thinking.

AI can now produce a convincing presentation in minutes. It can summarise a complex document, prepare an analysis and generate a recommendation without the person using it necessarily understanding the subject particularly well.

That is a significant change. Yet many performance systems still place considerable weight on visible output.

I'm not suggesting output no longer matters. It absolutely does. But we need to become much better at distinguishing between producing something and creating value.

The World Economic Forum's Future of Jobs Report 2025, based on a survey of more than 1,000 employers, provides an interesting perspective. Analytical thinking remains the most frequently identified core skill, cited by approximately seven in ten employers. The report also highlights the growing importance of resilience, flexibility, agility, leadership and social influence.

These are not simply production skills. They influence how people assess situations, respond to uncertainty, work with others and make decisions. Employers increasingly recognise their importance.

But recognising a skill as important and actually rewarding it are two different things. And that is where I think the problem starts.

What are we actually rewarding?

I've spent much of my career working with commercial teams, delivery teams and organisations where the expectation is ultimately to produce a result. One thing I've learned is that the person producing the most material isn't necessarily the person creating the most value.

Sometimes it's the person who asks the uncomfortable question before everyone commits to the wrong decision. Sometimes it's the person who understands what the customer actually needs, rather than simply responding to what the customer initially requested. Or the person who identifies a problem early enough to prevent it from becoming an expensive one. And sometimes it's the person who takes responsibility for getting something over the line when the original plan isn't working.

These contributions have always mattered. But they are not always easy to measure. It's much easier to count completed tasks than to assess the quality of someone's judgement.

And that creates an interesting contradiction. We can tell people that judgement, accountability and problem-solving matter, while continuing to assess them primarily on activity and output.

Belief is cheap. The scorecard is what shapes behaviour.

If you reward people for producing reports, they will produce reports. If you reward activity, you will get activity. And now AI makes it easier than ever to generate both. You automate the very thing your incentives still celebrate.

I think this is one of the more important management challenges AI is creating. Not simply deciding where to deploy the technology, but reconsidering how we assess the people working alongside it.

Changing the scorecard is harder than adopting AI

It's relatively straightforward to give employees access to AI tools. Changing how you measure performance is considerably more difficult.

Take a commercial team. You can measure calls made, meetings held, proposals submitted and opportunities created. Those metrics have their place. They help us understand whether the necessary activities are happening.

But they don't necessarily tell us whether someone understands the customer, has identified the right opportunity or is moving the business towards a commercially sound outcome.

The same applies to delivery teams. Completing tasks and meeting deadlines matter. But so do identifying dependencies, challenging unrealistic requirements, managing risks and making decisions that prevent problems later.

AI can help with many of these activities, including the analysis that supports good decisions. What it cannot automatically establish is whether the person using it has exercised sound judgement or accepted responsibility for the outcome.

That distinction matters. I would rather have someone who uses AI to prepare a shorter, better-informed recommendation and then takes responsibility for the decision than someone who produces twenty impressive pages without understanding what they mean.

But how many performance systems would consistently recognise that difference?

I don't think there is a simple answer. Judgement is difficult to measure objectively. So are ownership, influence and the ability to prevent problems before they occur.

We also need to avoid replacing one poor measure with another. A manager's subjective opinion of someone's judgement is not automatically a better performance metric than output.

The answer is likely to involve a combination of measurable outcomes, decision quality, customer impact, collaboration and accountability. And the balance will differ depending on the role. A processing environment still needs accuracy and throughput. A sales team still needs revenue. A delivery team still needs to deliver.

The point is not to stop measuring output. It's to stop confusing output with the full value of someone's contribution.

What happens when AI becomes good at those skills too?

There is an obvious counterargument. AI is already improving at analysis, negotiation support, planning and decision-making. Over time, it may become capable of performing more of the activities we currently associate with human judgement and leadership.

I think that's a fair challenge. It also means we should be careful about declaring any particular human skill permanently protected from automation.

Judgement, ownership and influence may become more valuable in certain roles as AI takes over routine activities. That doesn't mean they will remain scarce forever. The uncomfortable reality is that the capabilities organisations value will continue changing as technology improves. People will need to adapt, and so will the organisations employing them.

There is another important consideration. These changes will not benefit everyone equally.

Autor's research reminds us that technological progress can have uneven consequences for employment and wages. Automation has contributed to the decline of routine, middle-skill employment in advanced economies, alongside growth in higher- and lower-skill occupations.

We shouldn't assume that everyone whose tasks become automated will simply move into more rewarding work. Some people will find new opportunities. Others may struggle to make that transition.

Better performance management won't solve that problem on its own. Organisations will also need to think seriously about how they develop people and give them opportunities to build the capabilities that are becoming more valuable.

What this means for careers

I think there is a practical question every professional should be asking. Which parts of my work are becoming easier for someone else to reproduce using AI? And which parts depend on capabilities that remain difficult to replicate?

That doesn't mean abandoning the first category. Knowing how to use AI effectively is increasingly important. It can improve productivity and allow people to take on work that previously required more time or resources.

But simply knowing how to use AI is unlikely to remain a meaningful differentiator when everyone has access to similar tools.

The more interesting opportunity is what you do with the time and capability AI gives you. Do you develop a better understanding of your customers? Do you become more commercially effective? Can you make better decisions, identify opportunities others overlook or take responsibility for more complex outcomes?

These are questions worth asking whether you're starting your career or have been managing businesses for decades.

And for leaders, there is another responsibility. If we want people to develop these capabilities, we need to create environments that recognise and reward them. Otherwise, we shouldn't be surprised when people continue optimising for the old measures.

The pattern across this series

Over the past fifteen weeks, one theme has kept emerging. AI is making certain capabilities more accessible and reducing the effort required to produce work that previously demanded significant time and expertise. That changes where value is created.

This week adds another dimension. What an organisation says it values is not necessarily what its performance system rewards. I think that gap is going to become increasingly difficult to ignore.

A prediction I can be wrong about

By the end of 2030, I expect a majority of large employers to place greater formal weight on decision quality, accountability and problem-solving when assessing and promoting people, rather than relying primarily on activity and output volume.

The test will be whether employer surveys and publicly disclosed performance frameworks show that shift compared with 2026. If they don't, my prediction was wrong.

There are obvious challenges to measuring this consistently, particularly because internal performance frameworks are not always public. But I believe the direction of change is worth putting on record.

The bottom line

AI is changing the economics of work, but I think we're underestimating what that means for the way we manage people.

The challenge isn't simply getting employees to use AI. It's understanding how their contribution changes when they do. And then making sure our performance systems recognise the difference.

We should still expect people to deliver. We should still measure outcomes and hold people accountable. But we need to be much more thoughtful about what good performance actually looks like when producing the work is no longer the hardest part.

I'll leave you with two questions.

If AI became twice as capable tomorrow morning, which part of your contribution would become more valuable, not less?

And for those leading organisations, a more uncomfortable question: does your performance system actually reward that contribution, or are you still measuring the output AI can already produce?

---

Sources

Autor, D. H. (2015). "Why Are There Still So Many Jobs? The History and Future of Workplace Automation." Journal of Economic Perspectives, 29(3), 3–30. doi.org/10.1257/jep.29.3.3

World Economic Forum (2025). "The Future of Jobs Report 2025." weforum.org