Skip to main content
Abstract soft mint-green flowing forms suggesting cognitive flow and structure
Synthetic Minds at Work

AI as an Executive Function Prosthesis

On planning, friction and the redistribution of cognitive effort

March 28, 2026

There is a tendency to describe generative AI as a tool that enhances thinking. The phrase appears reasonable. It suggests augmentation, a broad increase in capability, a general improvement in how individuals engage with complex tasks. In many contexts, this is how the technology is experienced, as work that feels more manageable, ideas that are easier to articulate, decisions that appear less opaque.

Yet this description, while convenient, obscures something more specific. Generative systems do not enhance thinking in a uniform way. They intervene at particular points in the process through which thinking becomes work. More precisely, they alter the effort required to move from intention to action.

This distinction is easy to miss because the process itself is rarely examined.

Generative systems do not enhance thinking in a uniform way. They intervene at particular points in the process through which thinking becomes work.

The Invisible Transitions

Work is often described in terms of outputs, whether a report that has been written, a decision that has been made, or a plan that has been executed. The intermediate steps, the cognitive transitions that make these outcomes possible, remain largely implicit. They are, however, where much of the difficulty resides.

Beginning a task requires more than understanding what needs to be done. It demands the ability to establish a starting point amid ambiguity, imposing structure where none yet exists and determining what matters before anything else can proceed. This process is often treated as trivial, though the effort it actually requires is rarely acknowledged.

Continuing a task demands that coherence be maintained across interruptions, shifting priorities, and incomplete information, while context is held in mind and updated as conditions change against the constant pull of competing demands.

Completing a task involves translating internal reasoning into forms that others can interpret and act upon, a process that requires judgement about what to include and what to omit, how to frame uncertainty, and how to present conclusions that may still be provisional.

These activities constitute the work itself. What is commonly described as execution is, in practice, the visible surface of a deeper layer of cognitive orchestration, encompassing the planning, sequencing, prioritising, framing and translating that determine whether ideas remain internal or become actionable.

What is commonly described as execution is, in practice, the visible surface of a deeper layer of cognitive orchestration.

Unevenly Distributed Friction

These orchestration processes are also unevenly distributed. For some individuals, the transition from intention to structured action occurs with relative continuity. The steps are not effortless, but they do not require sustained negotiation. Attention aligns with priority often enough to maintain forward movement.

For others, the same transitions involve friction. The beginning of a task is delayed because a stable entry point refuses to materialise, even when the underlying problem is well understood. Structure must be assembled deliberately rather than emerging on its own, and attention requires continuous intervention to remain aligned with intention. The difficulty resides less in thinking about the problem than in organising that thinking into a form that allows the problem to be addressed.

This is where generative systems begin to exert their influence. They do not remove the need for thinking, but they change the cost of organising it. A task that previously required internal structuring can be approached through dialogue, with a sequence proposed externally, adjusted and refined. The question of where to begin, which can represent a disproportionate barrier, becomes less opaque when a system offers an initial frame.

This frame is rarely perfect, nor does it need to be, since its function lies in reducing the friction associated with entering the problem rather than solving it outright.

Generative systems do not remove the need for thinking, but they change the cost of organising it.

Prosthesis, Not Replacement

In this sense, generative AI begins to resemble something more specific than a general-purpose tool. It functions as a form of prosthesis, closer to compensation than replacement, supporting a function that may be unevenly reliable when carried out internally and providing an external structure that can be engaged with rather than assembled from nothing.

The analogy is imperfect, but instructive. A prosthetic device does not make an individual fundamentally different, yet it alters the conditions under which certain actions are possible and changes the relationship between intention and execution. Generative systems operate in a similar way at the level of cognition, making it possible for planning to occur outside the confines of internal processing and for sequencing to be negotiated through interaction rather than imposed internally, while thought itself becomes easier to translate into communicable form.

The result is a redistribution of effort rather than a uniform increase in capability. For individuals whose primary friction lies in the orchestration of tasks, this redistribution can be significant, as activities that were previously avoided or delayed due to their organisational complexity become accessible and the cost of initiating and maintaining structured work decreases. For others, whose processes were already aligned with the demands of internal structuring, the effect is more incremental, with work becoming faster and more efficient while the underlying mode of operation remains largely unchanged.

It functions as a form of prosthesis, closer to compensation than replacement, supporting a function that may be unevenly reliable when carried out internally.

The Asymmetry of Improvement

This asymmetry is often overlooked. If both individuals produce more output, the improvement appears equivalent, and what remains hidden is that the nature of the improvement differs considerably. For one person, the system optimises an existing capability, while for the other it alters the very feasibility of engaging with the task at all. As the first essay in this series explored, organisations have long rewarded the ability to present thought in a particular format, and AI is now collapsing the cost of that presentation in ways that affect different cognitive profiles unevenly.

The distinction becomes more pronounced when these systems are used not merely as generators of finished output but as environments for thinking itself. Questions asked not to receive definitive responses but to explore possibilities, partial structures externalised and then revised, ambiguity examined through iterative interaction rather than resolved immediately. This mode of engagement is not equally intuitive for all users, yet for individuals already accustomed to navigating internal variability and iterating between different frames of understanding, it may feel strikingly natural, resembling processes they have developed independently and now find supported externally. For others, the introduction of a system that produces plausible structures can create a tendency to accept those structures prematurely, resulting in a more limited use focused on optimisation rather than exploration.

For one person, the system optimises an existing capability, while for the other it alters the very feasibility of engaging with the task at all.

What Is Being Evaluated

These differences matter because they shape how the redistribution of cognitive effort translates into observable outcomes and complicate the interpretation of performance in ways that most evaluation frameworks are not designed to capture.

If one individual produces a well-structured plan with minimal apparent effort, and another arrives at a similar result through extended interaction with a system, the outputs may appear equivalent even though the processes that produced them are fundamentally different. In one case the structure emerges internally, while in the other it is co-produced through interaction, and neither process is inherently superior since they reflect different configurations of cognition and support.

For organisations, this introduces a question that is not easily answered. What, exactly, is being evaluated. If performance is assessed solely on output, the distinction between internally generated structure and externally supported structure becomes irrelevant. If, however, the process of arriving at that output carries significance, the role of the system cannot be ignored. This tension becomes particularly acute when knowledge itself begins to outpace the structures through which it is converted into decisions, a dynamic examined in the Decision Architecture series, and specifically in When Knowledge Outpaces Decision.

What, exactly, is being evaluated. If performance is assessed solely on output, the distinction between internally generated structure and externally supported structure becomes irrelevant.

A More Fluid Distribution

This tension is unlikely to be resolved through simple adjustments. It reflects a deeper shift in how work is organised. As generative systems become more integrated into everyday workflows, the boundary between thinking and tool use becomes less clear, with planning increasingly externalised and execution intertwined with preparation in ways that make the process of working altogether more fluid, along with the distribution of effort within it.

What emerges is a world in which the relationship between effort and output is less stable than it once appeared. Some forms of friction are reduced while others persist, and new forms of coordination become possible even as new dependencies are introduced.

The notion that AI simply enhances thinking fails to capture this complexity, suggesting a uniform effect where none exists. What is taking place is more specific, as the cost of organising thought is changing and with it the conditions under which different individuals can engage with work.

Questions this essay answers

What is meant by an executive function prosthesis?

Executive function refers to the cognitive operations that organise effort across time, including planning, prioritisation, working memory and self-monitoring. Generative systems perform a meaningful portion of those operations on behalf of the user, which makes them functionally equivalent to a prosthesis for the underlying capacity.

Does this benefit only neurodivergent employees?

The benefit is broadest for employees who previously paid the highest cost for organising their own cognitive output, which includes a substantial neurodivergent population. The same mechanism, however, lowers planning friction for the wider workforce, with effects that vary by role and seniority.

What is the operational risk that follows?

The operational risk is that organisations come to depend on the prosthesis without recognising the dependency, with the result that capability appears to be located in the employee when in fact it is distributed across the employee and the system.

Sources

  1. Barkley, R.A., Executive Functions: What They Are, How They Work, and Why They Evolved, Guilford Press, 2012. Foundational framework for understanding executive function as self-regulation.
  2. Diamond, A., "Executive Functions," Annual Review of Psychology, Vol. 64, 2013. Comprehensive review of the cognitive processes governing planning, sequencing and task initiation.
  3. Risko, E.F. and Gilbert, S.J., "Cognitive Offloading," Trends in Cognitive Sciences, Vol. 20, No. 9, 2016. Research on how external tools alter the distribution of cognitive effort.
  4. Heersmink, R., "A Taxonomy of Cognitive Artifacts," Review of Philosophy and Psychology, Vol. 4, 2013. Theoretical framework for understanding tools as cognitive prostheses.

About this Series

This article is part of the Synthetic Minds at Work series exploring how generative AI is reshaping the role of cognition in organisations.