The Architecture of Human Judgment
By Prof. Dr. Uwe Seebacher
Consider an imagined office. A report that once occupied an afternoon is finished before lunch. The document is checked, the numbers are sound, the client is satisfied. A small technological promise has been fulfilled.
Then the afternoon fills up.
Another report becomes possible. A meeting acquires a longer briefing. Someone asks for three alternative scenarios. By evening, the employee has produced more and recovered none of the time. The machine has kept its promise. The organisation has made a different decision about the proceeds.
This distinction belongs at the centre of the debate about artificial intelligence and work. Technical efficiency creates capacity. Institutions allocate it. Whether an hour becomes additional output, better service, learning, profit or personal freedom depends on decisions about goals, rewards and authority.
The question is therefore deceptively simple: who owns the time that AI saves?
The evidence makes room for two truths
There are sound reasons to welcome AI assistance. In a study published in The Quarterly Journal of Economics in 2025, Erik Brynjolfsson, Danielle Li and Lindsey Raymond examined its staggered introduction among 5,172 customer-support agents. Productivity, measured as issues resolved per hour, increased by 15% on average. Benefits varied across workers, with larger gains among less experienced agents. The study also found improvements in aspects of customer interaction. Its setting was one firm; its findings do not establish the economy-wide consequences of AI.
These gains matter. A difficult task can become more manageable. An inexperienced worker can gain useful support. Customers can receive a better answer. Treating every increase in productivity as a disguised injury would erase benefits that deserve to be understood.
Alongside this evidence sits a different observation. In a UC Berkeley research interview published on 18 February 2026, Xingqi Maggie Ye described research with Aruna Ranganathan: an eight-month ethnography at a technology company of about 200 people, including more than 40 interviews. Employees expanded their task range, carried work into former pauses and kept more activities running in parallel. This was a qualitative study of one workplace and was described as research in progress. It identifies a possible mechanism of work intensification; it cannot tell us how common that mechanism is across society.
The findings concern different settings and methods. Read together, they invite a question that a productivity figure cannot settle: what happens to the working day after an individual task becomes faster?
How a possibility becomes an expectation
One possible answer begins with an ordinary organisational habit. Yesterday's exceptional performance becomes tomorrow's baseline.
Once a team demonstrates that it can deliver a document in an hour, planning may begin to assume an hour. The time released from drafting is allocated before anyone discusses its purpose. More versions become affordable. More requests become reasonable. The employee who previously appeared conscientious can begin to appear slow while working at the same intensity.
This is a structural interpretation, rather than a claim that every employer behaves this way. It describes a mechanism worth examining: a new capability enters the organisation as an option and returns to the individual as an obligation.
Sometimes no manager needs to issue an instruction. Colleagues respond to one another. Clients adjust their expectations. Professionals take pleasure in discovering what they can accomplish. Competition turns voluntary experimentation into a standard that becomes difficult to refuse.
There is genuine satisfaction in extending one's abilities. The problem arises when a person loses a meaningful say over the extension. A demanding project freely chosen and a permanent demand for availability can occupy identical hours while producing very different experiences of work.
Methods and structures become decisive at this point. They determine who can commit another person's time, how priorities are revised and whether an improvement in speed automatically changes the expected volume of work.
The missing unit of measurement
A task has a beginning and an end. A working life has dependencies.
Measuring a faster draft tells us something useful about that draft. It says less about the time spent verifying it, coordinating its use, correcting downstream misunderstandings or responding to the additional requests it enables. These activities may be small or substantial. Their size needs to be observed.
An organisation can therefore become more efficient at a visible activity while creating work elsewhere. A shorter preparation phase might be followed by a more demanding review phase. Responsibility can remain with the human being even when the production of the first answer becomes almost effortless.
The relevant boundary is the complete process. Where does work begin? Where does it finish? Whose effort is counted? A gain recorded in one department can be partly paid for by another department's attention. A manager's saved hour can arrive as several colleagues' new obligation.
For this reason, I would distinguish three questions in any serious assessment: how much capacity has been created; where that capacity has been allocated; and how much discretion people retain over their time. These are proposed analytical distinctions, not an empirically validated index.
They also change what success looks like. A process might produce more with the same working hours and improve service. Another might preserve output while reducing hours. A third might protect time for learning or difficult judgment. Each can be valuable. The choice deserves to be explicit.
The old question inside the new technology
In Economic Possibilities for our Grandchildren (1930), John Maynard Keynes imagined how technological and economic progress might alter the necessity of work over a hundred-year horizon. His discussion of much shorter working time was accompanied by a more difficult question: how would people learn to use greater freedom well?
That question survives the historical assumptions of the essay. A society can become accomplished at producing the means of life while struggling to decide what a good life should contain.
Efficiency offers a comparison between resources and results. It cannot, by itself, tell us which results deserve pursuit. The value of an uninterrupted conversation, an unhurried act of care or an afternoon spent thinking depends on purposes that a speed measure does not supply.
Nor is freedom exhausted by leisure. People may choose to devote released time to a craft, an enterprise, teaching or demanding research. The philosophical issue concerns their capacity to participate in that choice. A fuller life can include more work. Its fullness is difficult to infer from the number of tasks completed.
There is also a question of distribution. Some workers can negotiate protected time; others face insecure employment, relentless demand or tasks that technology barely accelerates. Consumers may benefit through lower costs or improved services. A serious discussion must examine these different claims on the gain, rather than assume that every saved minute can simply become private leisure.
Designing a place for the saved hour
For leaders, this suggests a practical discipline: decide how gains will be evaluated and allocated before a pilot quietly changes the baseline.
A team can begin with a bounded process and a clear account of the outcome it wants to preserve. During the pilot, it should record the whole workflow, including review and coordination, and ask the people doing the work where demands have moved. Output and quality belong in that record, together with working hours and employees' reported discretion.
The next decision is allocation. Some capacity may be needed to improve service or absorb a backlog. Some may support training, reflection, recovery or shorter hours. There is no universal proportion. There should be a visible decision, a stated reason and a way for affected people to challenge the assumptions.
Crucially, the new baseline should remain open to review. A temporary burst of experimentation is a weak basis for a permanent workload expectation. A useful working hypothesis would be that teams which make allocation explicit retain more discretion than comparable teams which let expectations adjust informally. Testing that proposition would require a suitable design, attention to differing workloads and a definition of discretion agreed in advance.
This connects with the question behind my work on Lazy Excellence: how can clarity about what matters improve the way we work? Here, that concern extends from individual choices to the structures that make those choices possible. A person may learn to prioritise with great skill and still inhabit a system that treats every available minute as unclaimed production capacity.
Leadership becomes visible in the treatment of that minute. Does it disappear into a new target? Does it support a better decision? Can its use be discussed?
Return to the imagined office. The report is finished. The cursor blinks. An afternoon has become possible.
What happens next will reveal what the organisation believes time is for.
For editors and interview producers: For conversations about AI, work, organisational structures and human judgment, contact us@uweseebacher.org.
Author's note: This essay develops an interpretive and normative argument. Proposed analytical distinctions and practical steps are not presented as a validated intervention. The opening office scene is hypothetical. Lazy Excellence is a book by the author; the link provides related context.
Sources
- Brynjolfsson, E., Li, D. & Raymond, L. (2025). Generative AI at Work. The Quarterly Journal of Economics, 140(2), 889–942.
- Counts, L. (18 February 2026). Does AI Actually Free up Workers' Time? UC Berkeley research interview with Xingqi Maggie Ye, discussing work with Aruna Ranganathan.
- Keynes, J. M. (1930). Economic Possibilities for our Grandchildren. The Nation & Athenaeum; subsequently collected in Essays in Persuasion.
