Artificial intelligence presents organisations with a curious paradox. The technology can draft documents, analyse information, generate code and respond to customers in a fraction of the time these activities once required. Yet executives investing heavily in these capabilities are still asking a basic question: where is the corresponding economic value?
There is less of a contradiction here than it appears.
Technology vs company
The problem is that AI makes tasks more productive, while organisations create value through systems of interdependent tasks. Making individual pieces of work dramatically faster does not necessarily make the organisation containing them more effective. PwC's January 2026 survey of 4,454 chief executives found 56 per cent report no financial benefit from AI.
We have seen a version of this before. In 1987, economist Robert Solow observed that computers were visible everywhere except in the productivity statistics. The gains from information technology became clearer only as organisations learnt to redesign work around its possibilities. The lesson was not simply that new technologies take time to generate returns. It was that technological change must be accompanied by organisational change.
AI may be creating the same challenge at considerably greater speed, but organisations are redesigning tasks faster than they are redesigning themselves.
The mismatch
Consider three levels at which this mismatch occurs.
First, organisations need to rethink what they are trying to achieve with AI. The most obvious business case is efficiency. If AI allows work that previously took three hours to be completed in 30 minutes, calculate the resulting cost saving.
However, efficiency and value creation are not the same thing. As similar AI capabilities become widely available, doing a standardised activity cheaply may quickly cease to provide differentiation. The more interesting question is what an organisation can now do that was previously too expensive, too slow or too demanding of scarce expertise.
This distinction is particularly relevant in the UAE. In an economy pursuing ambitious growth and diversification while competing for skilled talent, the opportunity from AI cannot simply be about reducing labour. It can also be about extending scarce expertise: evaluating more opportunities, serving more customers or providing services that previously would have been uneconomic. UAE AI skills job postings have tripled since 2021 and wage premiums have reached 92 per cent.
Second, organisations need to redesign how work fits together.
Most AI implementation begins, quite reasonably, by decomposing work: Which tasks can AI perform? Which bottleneck can it remove? Where can time be saved? But decomposition creates another problem: someone has to put the work back together.
Imagine a professional workflow in which AI conducts the research, another system performs the analysis and a third prepares the final output. Each component may become faster. Yet the workflow as a whole can become more fragile.
The reason is that apparently inefficient human activities often perform functions that are easy to overlook. People do not merely execute tasks. They catch unusual cases, question assumptions, connect information across stages and take responsibility when something goes wrong.
This is why the crucial question is not simply, “Can AI do this task?” It is: “If AI does this task, who can verify its output, and who remains accountable?”
Where AI outputs can be cheaply verified, workflows should be redesigned to make verification systematic: evidence alongside factual claims, tests alongside code, audit trails alongside decisions. Where outputs cannot be reliably checked by a rule, human judgment remains essential.
This is more useful than the generic prescription to keep a human in the loop. The objective should be to put humans in the loops where judgment and accountability are required – and redesign the other loops so that they are safely verifiable.
Third, organisations need to rethink the role of the human worker.
The usual approach is based on comparative advantage: allocate to AI the tasks it performs better and leave humans with whatever remains. But that risks defining human work as a shrinking residual.
A better model is to think of a role as a person responsible for a larger unit of output, supported by AI agents that execute parts of the work. Two human capabilities then become especially important: problem formulation – deciding what problem actually needs solving, and solution validation – judging whether an AI-generated answer is good enough to act upon.
This shift also changes how organizations think about human capital. The challenge isn't just equipping people with new technical skills - it's rethinking how roles, capabilities and learning evolve as AI becomes embedded in everyday work. It's a question we've been exploring closely with leaders through INSEAD's Human Capital in the Age of AI programme”.
There is also a longer-term risk. If organisations automate the experiences through which junior employees historically learnt to exercise judgment, they may increase today's efficiency while weakening tomorrow's supply of expertise. Another study in May found junior employment fell roughly 9 per cent at AI-adopting firms.
Productivity paradox
The AI productivity paradox is therefore fundamentally an organisation-design problem. Closing the gap requires redesign at three connected levels: what the organisation seeks to accomplish, how its workflows combine human and AI contributions, and what people ultimately remain responsible for.
As AI becomes more capable, organisation design, therefore, becomes more important, not less. The easier it becomes to decompose work and give its pieces to machines, the more important it becomes to decide how those pieces should be recombined into an accountable whole.
The organisations that capture the most value from AI may not be those that adopt it fastest, but those that redesign themselves around it most intelligently.
Phanish Puranam is the Roland Bergerchaired professor of strategy and organisation design at Insead.
