Your Job Description Is a Fossil

Your Job Description Is a Fossil

How AI Changes the Structure of Work Through Distributed Cognition

September 22, 2026

Jiajie Zhang, PhD

Abstract

Debates about AI and employment usually begin with a fixed list of tasks and ask which machines can perform. This essay argues that the list itself changes. Anthropic’s illustration of nursing work offers a compact example: some activities remain human, others are augmented or automated, and new activities emerge. Distributed cognition explains the deeper mechanism. Tasks depend on the relationships among people, representations, technologies, and institutional arrangements. As AI becomes a representational medium that can generate, analyze, and revise information, it changes both the distribution and the structure of cognitive work. The implications extend beyond automation to professional expertise, education, institutional design, and the allocation of economic gains. Institutions must therefore reconsider how work is organized, which responsibilities remain essential, and which exist because earlier systems could do no better.

1. The Job Description as a Historical Arrangement

We are debating which jobs AI will eliminate while treating the current organization of work as if it were a law of nature.
It is a historical arrangement. Your job description records what people had to do, with the technologies they had, inside the institutions they inherited. Some of those responsibilities express enduring human needs. Others exist because information was difficult to find, expertise was scarce, or coordination was expensive.
AI puts that arrangement up for renegotiation.
A beautiful illustration in Anthropic’s economic scenario explorer (Anthropic, 2026) makes this visible through a nurse’s work. Some activities remain entirely human. Others are augmented or automated. New activities appear, including evaluating AI triage and reviewing a proposed care plan. The figure illustrates possibilities within a model exploring economic scenarios through 2030.

Figure 1. A nurse’s tasks in Anthropic’s economic scenario explorer. Source: Anthropic (2026). The figure illustrates possible changes in task composition.

I read the figure as a compact illustration of distributed cognition, the framework at the center of my book, The Cognitive Revolution: How AI Is Reorganizing Intelligence, Expertise, and Institutions (Zhang, 2026b).
The most consequential feature is easy to overlook: the list of tasks itself changes. Anthropic explicitly recognizes this. Distributed cognition helps explain why it happens.
That observation should unsettle anyone whose AI strategy consists of taking today’s workflow and deciding which boxes a machine can fill.

2. Distributed Cognition Changes the Unit of Analysis

A profession has never been simply a collection of abilities inside one person’s head. A nurse works through relationships with patients and colleagues, records, instruments, protocols, schedules, and organizational routines. What the nurse can notice, remember, infer, and accomplish depends on how those elements work together.
That is what distributed cognition asks us to examine: how thinking is accomplished across people and the representations they use, through successive interactions over time. It gives us a larger unit of analysis than the individual mind.
Consider the difference between remembering a patient’s changing condition and seeing it represented as a trend. The representation makes a pattern available for inspection. It changes the mental operations required to recognize that pattern. It also introduces choices about what gets measured, what gets displayed, and what remains invisible.
The information’s form helps determine the work required to use it. A task is partly a product of the medium in which we attempt it. Change the medium, and you can change what the task requires, who can perform it, and whether it needs to exist at all.

3. AI as a New Representational Medium

This is why the phrase “AI tool” can be so intellectually limiting. It encourages us to picture an unchanged professional performing an unchanged task with a more powerful instrument. That picture captures part of the transition. It misses what happens when the environment in which work is formulated, represented, and coordinated begins to change.
An AI system can generate a candidate explanation, compare alternatives, transform a record into a plan, or revise a proposal when a constraint changes. The representation becomes something we can question and ask to transform itself. Its outputs can then shape what we notice and do next.
We are building a medium that participates in the work of thinking.
In “Beyond AGI”, I describe this process as the computationalization of distributed intelligence (Zhang, 2026a). AI brings more of the cognitive work previously performed through interactions among people, documents, tools, and institutions into active computational processes. Here, the consequence is a changing structure of work: both its allocation and its content become open to redesign.

4. Tasks Change When Their Representations Change

To see the practical difference, imagine a ward’s staffing schedule. A static roster displays who is assigned where. An AI-supported planning environment could bring together staff availability, qualifications, patient needs, workload, and changing constraints, then propose alternative arrangements.
Scheduling has always required people to weigh constraints, negotiate exceptions, and recognize needs that a roster cannot capture. AI changes the distribution of that work. Generating candidate schedules can move into the system, while specifying its goals, challenging its assumptions, and managing exceptions take on a different role. The task is reorganized around a representation that can propose and revise arrangements.
Suppose the proposed schedule is mathematically efficient but ignores the exhaustion of a particular nurse, a patient’s need for continuity, or the informal support an experienced colleague provides. The output may look excellent precisely because the representation has left out something essential.
Professional expertise now includes detecting that omission and knowing how to correct it. Domain knowledge becomes critical at a different point in the process. A polished output can conceal a badly framed problem.
Dividing existing tasks between humans and AI is only the beginning. Changing the representation can change the task’s structure, the questions people ask, and the standards by which an answer should be judged.
Even an activity that remains entirely human can acquire a different role within the larger system. A conversation with a patient may become more consequential when it reveals something that no dashboard captured. A colleague’s objection may expose a weakness that every automated check missed.
The consequences travel through the relationships among tasks. Coloring one box “human” does not isolate it from changes elsewhere.

5. New Tasks Emerge from New Possibilities

The new work also extends beyond supervising AI.
Imagine a research claim that remains connected to its supporting data, competing explanations, and tests that could overturn it. An AI system could help keep those connections current, surface contradictory findings, and propose analyses when new evidence arrives. A research team could organize work around maintaining an argument that remains open to interrogation after publication.
That creates responsibilities for deciding what triggers reconsideration, tracing how a conclusion changed, and resolving disagreements between successive analyses. The ingredients have precedents. Their integration makes a different object of work practical: a claim with an ongoing life that someone must maintain and govern.
This is a design possibility extending beyond Anthropic’s example. New tasks need not be activities no human has ever imagined. They can emerge when familiar activities become newly connected, continuous, or feasible at scale. The goals themselves expand along with our capacity to pursue them.
The future task list cannot be discovered simply by auditing the present one.

6. Institutional Leadership Becomes a Design Responsibility

This is where institutional leadership becomes decisive.
A hospital can install powerful AI and retain fragmented handoffs. A university can give everyone a chatbot and preserve a curriculum organized around producing answers that are now readily available. A research organization can generate more proposals while leaving its processes for selecting, testing, and learning from ideas untouched.
All three can report impressive adoption numbers while preserving the organizational constraints that made the work cumbersome in the first place. An institution can become very good at using AI to maintain an obsolete design.
Institutions need to decide how work should be organized when cognition can be distributed differently. That means examining who frames a problem, which evidence enters the system, how competing interpretations are challenged, who can act, and how mistakes lead to learning.
It also means giving people the authority and resources to exercise the judgment we keep claiming will become more important. Calling someone a “human in the loop” accomplishes very little if that person has thirty seconds to approve an output, cannot inspect its basis, and has no practical power to reject it.
We can distribute cognitive work widely. Responsibility still needs a name, an owner, and the power to intervene.

7. Education Must Prepare People to Redesign Work

Education faces the same design problem. Students need deep knowledge to recognize a consequential error, formulate a worthwhile question, and challenge a misleading representation. They also need to learn how to work within, evaluate, and improve systems that extend beyond their own minds.
An assessment that measures only what a student can produce alone tells us too little about that competence. An assessment that accepts whatever a student produces with AI tells us too little as well. We need to examine the quality of the student’s reasoning, the contribution of the surrounding system, and the student’s ability to explain and defend the result.
Teaching yesterday’s task bundle more efficiently will not prepare students to redesign tomorrow’s work. A curriculum becomes a liability when it confuses the current division of labor with the permanent structure of a discipline.

8. Who Benefits from Reorganized Intelligence?

There is an uncomfortable economic point here, too. New tasks do not promise enough new jobs, equivalent wages, or a fair distribution of gains. Distributed cognition is an explanation of how work can be reorganized. It offers no automatic protection against displacement or concentrated power.
If an institution saves time, leaders decide whether that time becomes better service, more learning, greater workload, or fewer positions. If a professional’s expertise becomes embedded in a shared system, ownership and access determine who benefits from its wider use.
Those choices deserve to be visible. A productivity gain cannot tell us what an institution ought to do with it.
The most revealing question for an institution is therefore uncomfortable: which parts of our work exist because they serve a human purpose, and which exist because our old systems could do no better?
Answering it requires leaders to reconsider the cognitive architecture of the institution: how people, representations, and technologies make thinking and action possible. The organization chart, curriculum, and job description all become open to redesign.
A job description records an earlier answer to how work should be organized. It should never be allowed to veto the next one.

References

Anthropic. (2026). What will our economic future look like? Economic scenario explorer, version 1.0.
Zhang, J. (2026a, August 3). Beyond AGI: How frontier models are computationalizing distributed intelligence. dcognition.ai.
Zhang, J. (2026b). The Cognitive Revolution: How AI Is Reorganizing Intelligence, Expertise, and Institutions. Open Intelligence Press.

Continue Reading

This framework is based on my book, The Cognitive Revolution: How AI Is Reorganizing Intelligence, Expertise, and Institutions.
Book: https://www.amazon.com/dp/B0GWMHQSNG
Website: https://www.dcognition.ai

Jiajie Zhang, PhD
Author of The Cognitive Revolution: How AI Is Reorganizing Intelligence, Expertise, and Institutions
Dean and Glassell Family Foundation Distinguished Chair
D. Bradley McWilliams School of Biomedical Informatics
UTHealth Houston

How This Essay Was Created

I originate the ideas, arguments, conceptual framework, conclusions, and the first draft of every essay. I then use AI as a cognitive partner to challenge my thinking, improve organization, strengthen clarity, and refine the writing. The intellectual contributions are my own. This collaborative process reflects the central thesis of The Cognitive Revolution: intelligence increasingly emerges through distributed cognitive systems.

© 2026 Jiajie Zhang. All rights reserved.
Brief quotations with attribution and a link to the original publication are welcome. No part of this article may be reproduced, republished, or distributed in whole or in substantial part without prior written permission from the author.