The End of the Job Description

Hello! 👋

It’s Thursday, 1^st October 2026. Hello and welcome back to Bold Efforts!

A job description is an interesting artifact. It takes a collection of activities, assigns them to one person, and calls the resulting bundle a job. A marketing manager might be responsible for customer research, campaign strategy, content creation, agency coordination, and performance reporting. These activities require different skills, but we group them together because, historically, it has been convenient to employ one person to perform them.

The modern job is therefore not necessarily the most efficient way to organize work. It is a practical response to the economics of human labor and coordination. Hiring people is expensive, their time is finite, and dividing every activity among specialists introduces considerable overhead. Organizations solve this problem by creating roles with reasonably coherent sets of responsibilities.

AI (or SI) changes some of these assumptions. When software can perform individual activities at a fraction of their previous cost, the logic of bundling them into permanent jobs deserves reconsideration.

Consider a financial analyst whose responsibilities include collecting data, building models, preparing presentations, and advising management. AI can increasingly assist with the first three activities, although the extent of automation depends on the complexity of the work. The fourth involves contextual judgment, accountability, and an understanding of organizational priorities.

The conventional approach to AI adoption is to give the analyst better tools. The more fundamental approach going by first principles is to ask whether the original combination of responsibilities still makes sense.

This distinction reveals three possible outcomes for any activity. Some can be automated, others can be augmented, and a third category can be eliminated altogether. The last category is often overlooked.

If an executive can obtain reliable operational information whenever required, the weekly management report may no longer serve its original purpose. Automating its preparation would preserve an activity whose necessity has disappeared.

The important question is not simply how AI can help someone perform their job, but whether that job should contain the same activities in the first place.

This also challenges a common assumption about productivity. If AI reduces the time required for an activity by 80%, companies might expect the employee to perform five times as much of it. Yet that calculation assumes demand for the activity is unlimited and that the activity itself remains valuable. Neither assumption is necessarily true.

Most organizations follow a familiar hierarchy: departments contain teams, teams contain roles, and roles contain activities. Budgets, performance reviews, and career progression are built around this structure.

AI introduces the possibility of organizing more work around outcomes rather than predetermined roles. A small product team, for example, might combine human expertise with shared AI capabilities for research, prototyping, analysis, and operational execution. Rather than maintaining separate human capacity for every supporting activity, the organization could concentrate its employees on decisions, relationships, and specialized problems.

This does not imply that traditional departments will disappear. Organizations need continuity, accumulated knowledge, and clear accountability. Moreover, many activities are interdependent and cannot be separated without losing important context.

Nevertheless, the optimal boundary of a job may change. A company might need fewer people performing routine execution while placing greater value on individuals capable of directing complex systems. The managerial challenge shifts from allocating human labor across tasks to determining the most effective combination of human and machine capabilities.

There is an important economic consequence. Historically, increasing output often required increasing headcount. If technology allows output to grow without a proportional increase in employees, organizational scale becomes less closely associated with organizational size.

Recruitment today typically begins with an approved position. A hiring manager creates a list of responsibilities, identifies the required experience, and evaluates candidates against that combination.

This process can produce peculiar results. A candidate might possess exceptional judgment in the most commercially important part of a role but lack experience in several routine activities. Another candidate might satisfy every requirement without being particularly exceptional at anything.

The first candidate may be rejected because the job description rewards completeness rather than the capabilities that create disproportionate value.

As AI absorbs parts of existing jobs, companies may discover that their real requirement is not another marketing manager or strategy analyst, but someone capable of making particular decisions or exercising a form of judgment their existing systems lack.

Recruitment could therefore become more focused on identifying missing capabilities rather than replacing positions. This would also weaken the informational value of conventional career histories. A previous job title tells us what bundle of responsibilities someone was assigned, not necessarily which capabilities they developed.

Compensation presents a related challenge. Two employees with identical titles might generate vastly different outcomes because one can effectively orchestrate technology while the other relies primarily on personal execution. Traditional salary benchmarks, built around relatively standardized job descriptions, may become less informative.

Perhaps the most consequential implication concerns how people develop expertise.

Many professions depend on a progression from structured activities to complex judgment. Junior lawyers review documents before advising clients. Financial analysts construct models before making investment decisions. Engineers solve relatively contained problems before designing complex systems.

These entry-level responsibilities serve two purposes. They produce useful work for employers while providing training for employees.

AI threatens to separate those functions. An organization may discover that it no longer needs as many junior employees to perform routine activities, while still requiring experienced professionals capable of exercising sophisticated judgment.

This creates a potential long-term problem. If companies eliminate the work through which professionals traditionally acquire experience, where will the next generation of experts come from?

The answer cannot simply be to teach everyone to supervise AI. Effective supervision often requires a sufficiently deep understanding of the underlying activity to recognize errors and challenge assumptions. Someone who has never constructed a financial model may struggle to evaluate one whose formulas are technically correct but whose assumptions are economically unreasonable.

Organizations will therefore need to distinguish between activities that are merely expensive to perform and activities that are valuable precisely because performing them develops expertise.

The productivity gains from eliminating junior work could come at the expense of the future supply of senior judgment.

There is also a question of professional identity. People tend to describe themselves through occupations: designer, consultant, engineer, accountant. These labels provide social recognition and establish expectations about competence.

But occupations are themselves evolving bundles of activities. A designer who delegates much of the production process to AI may spend more time defining problems, evaluating alternatives, and making aesthetic judgments. An engineer may spend less time writing individual components and more time designing systems and verifying their behavior.

Their professions do not necessarily disappear. Instead, the activities that define competence within those professions change.

This could make careers less dependent on mastering a fixed set of tasks and more dependent on developing capabilities that remain valuable as tools evolve. Judgment, domain understanding, and the ability to formulate worthwhile problems may become increasingly important.

Yet these capabilities are not independent of technical knowledge. Good judgment generally depends on knowing enough about a field to distinguish a plausible answer from a correct one.

The job description is unlikely to disappear. Organizations still need accountability, predictable responsibilities, and contractual clarity. Employees also benefit from stability and a recognizable professional identity.

But the process through which jobs are designed may need to change.

Instead of beginning with an existing position and asking how AI can improve its productivity, organizations could begin with an outcome and work backward.

For much of the industrial and information age, the job has been the fundamental unit through which companies organize productive activity. AI makes it increasingly feasible to examine the activities underneath that unit and rearrange them. This is similar to redesigning workflows that I do for various enterprises.

We have spent decades fitting technology into jobs designed for humans. The next challenge is designing jobs for a world in which humans are no longer the only ones performing the work. Thank you for reading.

Best,
Kartik

I write Bold Efforts every week to think clearly about where work and life are actually headed. If you want these essays in your inbox, you can subscribe here.