If AI Multiplies Your Productivity, Who Keeps the Benefit?

For years we've been sold AI as the next great productivity revolution. And it probably is. Developers write code faster. Analysts produce documentation in minutes. Designers create prototypes in hours. Teams are able to produce more with less manual effort. So far, it all sounds like good news.

Yet every time I hear about "AI-First" or "AI-Centric" organizations, an uncomfortable question comes to mind: if productivity skyrockets thanks to AI, who actually benefits from that productivity?

Because the more I observe how AI is being adopted in companies, the more I'm convinced the debate is not about whether AI will replace workers. The real debate is how the value it generates will be shared.

The original promise

The initial narrative was simple. AI was going to eliminate repetitive tasks:

• Boilerplate code.
• Routine documentation.
• Writing tests.
• Information searches.
• Administrative tasks.

The logical consequence seemed obvious. If a person takes less time to do a task, that person should have more time available for higher-value activities.

More productivity. Less repetitive work. Better quality of life. At least on paper.

What seems to be happening

However, in many organizations the reality is somewhat different. AI is not only eliminating tasks: it is also expanding the scope of roles.

The developer no longer just develops. Now they also:

• Define specifications.
• Prepare context for AI models.
• Validate automatically generated results.
• Work closer to the business.
• Review generated documentation.
• Make decisions that used to fall to other specialists.

And this doesn't only affect developers. The same is starting to happen with analysts, product managers, architects, designers, and other profiles. AI is breaking down certain barriers between roles.

The silent merging of professions

For years the industry evolved toward specialization. We had teams full of experts:

• Product Owners.
• Architects.
• QA.
• Developers.
• Cloud Engineers.
• UX Designers.
• Security specialists.

Each contributed a specific piece of the process. Now another trend seems to be emerging: smaller teams, fewer dependencies, more autonomy, more distributed responsibility, and people able to cover more areas thanks to AI-powered tools.

The promise is attractive: less bureaucracy, fewer handoffs, more speed. But a reasonable doubt arises:

Are we eliminating work, or are we simply redistributing that work among fewer people?

Productivity doesn't mean lower cognitive load

There's something important that rarely shows up in corporate presentations: productivity and cognitive load are not the same thing. I may write fewer lines of code today than three years ago, but I may also have to manage more context, more tools, more decisions, more responsibility, and more different disciplines.

There's a curious asymmetry in how we measure work: we precisely quantify what goes down (time per task), but almost nobody measures what goes up: the context to manage, the responsibilities, the decisions, the tools. And they're equally important variables.

AI reduces operational effort, but in many cases it increases the complexity of the work. It's the same pattern I explored in "AI Cognitive Fatigue" and "How AI Is Reconfiguring Our Attention". And that creates an interesting paradox: a person can be more productive than ever and, at the same time, feel more tired than ever.

Money starts telling a curious story

A few weeks ago, a colleague fixed a duplicated-lines bug in two days that would have taken him a week and a half before. Same person, same codebase. The only difference: he now works with AI.

His reaction wasn't celebration. It was discomfort. "It makes me feel a bit dumb," he told me. "The cognitive context is volatile and the workload is still huge."

And he's absolutely right. Because a week and a half fighting a bug wasn't wasted time: it was absorption time. In those days you soak up the system, you understand why the code is the way it is, you map the limits of the inherited logic. Fixing it in two days with AI gives you the solution, but it doesn't give you that knowledge. The model does the diagnosis; nobody does the learning. Before, a week of investigation left a deep understanding of the system. Now, two days leave a solution. The uncomfortable question: where did the learning go? I explore this further in "Solving More, Learning Less".

Doing twice as much for the same pay. It's a 2-for-1 deal: we put in the work and the company keeps the product. The point isn't that the company captures part of the benefit (that's normal). The point is whether that benefit is being shared in a balanced way among everyone who contributes to generating it.

Because that "doing twice as much" isn't limited to a single task. AI doesn't just speed up what we already did: it absorbs tasks from other roles. Before, every piece of the process had its specialist: the architect designed the solution, QA validated, the documenter wrote, the analyst talked to the business. Now all of that converges into the same person.

Developer, architect, validator, documenter, and analyst in a single person... with the same compensation. With an important caveat: they don't literally do the work of five people. There used to be five clear owners for those decisions; now a single person oversees more parts of the process, AI-assisted to be sure.

Let's look at it with a simplified model and purely illustrative figures.

It can look like a good outcome: productivity increases significantly and nobody sees their compensation reduced. However, the picture changes when we add to the equation the new responsibilities, decisions, and cognitive load that come with that productivity. And if the company chooses not to grow, that same productivity means it needs fewer people for the same output: people become surplus. One out of every two old roles is no longer needed. More work for the same pay, or simply fewer jobs; either way, it's the other side of the 2-for-1 deal.

Before AI

A company has a developer whose total cost we set as an index of 100.

After AI

Introducing AI tools raises the total cost to 125, but productivity goes from 100 to 180.

The new invisible tax: tokens

There's something fascinating about this new economy. The extra productivity no longer just generates revenue for the company: it also generates revenue for AI providers. The more the worker produces, the more prompts, more queries, more generation, more tokens.

The economic chain looks like this: worker → AI → company. The worker generates more value, the company increases results, and the AI provider increases consumption. Both capture part of the benefit. The only actor whose compensation doesn't change automatically is the worker.

The business counterargument

To be fair, there's also a reasonable view from the company's side. The organization is the one buying the licenses, investing in the transformation, taking risks, training the teams, and maintaining the infrastructure. There's nothing unfair about that. And there are cases where the surplus gets reinvested in people: IKEA retrained 8,500 customer service employees into premium interior design advisors with its Billie chatbot, and the new service generated €1.3 billion in 2024. The problem appears when the relationship starts to become unbalanced.

The question nobody wants to answer

Imagine that in five years a professional is able to generate twice as much value, work with advanced AI, understand the business better, make more decisions, and have full ownership of their solutions. Would it still be reasonable to pay them exactly the same as when they did a single specialized function?

I don't have the answer. But I believe it's a conversation we're not having enough of.

Maybe we're asking the wrong question

Much of the public debate revolves around one question: will AI replace workers? Personally, I don't think that's the most interesting question.

The question that really worries me is another one: if AI multiplies human productivity, how will that increase in value be shared among workers, companies, and technology providers?

Because economic history is full of technological revolutions: the steam engine, industrialization, computing, the internet, the cloud, and now AI. Technology has always increased productivity. The discussion has never been whether productivity increases, but who captures that benefit. And I suspect that will be one of the great debates of the next decade.

Conclusion

AI is changing not only how we work, but how the value of that work is distributed. Before celebrating productivity, it's worth asking for whom it is actually productive.

If AI multiplies our productivity but also multiplies our responsibilities, are we facing a revolution that frees the worker, or simply a new way of concentrating more work into fewer people?