When AI makes answers cheap, work shifts toward questions and judgment

The familiar question is whether AI will take all the jobs. Nathaniel Whittemore's August 23 episode of The AI Daily Brief starts somewhere more useful: AI changes the work people can imagine, the way companies coordinate, and the skills that deserve attention. The 30-minute episode reads a selection of Every's Thesis Statements, a project in which builders and thinkers describe what work might look like after automation. 1

The episode's central idea is simple: when AI makes routine execution and answer production cheap, the scarce parts of work move toward choosing worthwhile questions, supplying live context, judging quality, coordinating people and agents, and taking responsibility for outcomes. The essays are predictions rather than labor-market measurements. Their value comes from making the mechanisms of change specific enough to examine.

Anne-Laure Le Cunff, founder of Ness Labs, argues that abundant answers raise the value of deciding which questions deserve attention. Her essay describes a work cycle built around experiments: a person frames a problem, chooses evidence that could change the person's view, lets AI compress the execution, and interprets what happened before choosing the next question. 2

That shift changes what a good worker contributes. A prompt is only the visible surface. Underneath it sit assumptions about the customer, the desired outcome, the acceptable risk, and the evidence that would count as success. Yash Tekriwal, an educator and go-to-market engineer at Clay, describes this underlying skill as computational thinking: breaking a messy goal into steps, defining inputs and outputs, and testing edge cases. AI makes those habits relevant to people who have never written code. 3

Dan Shipper, CEO of Every, makes the same shift from another angle. He argues that AI commoditizes the part of expertise that can be made explicit enough to train on. Default output becomes easier to obtain, while work that depends on a specific person, customer, codebase, or moment still needs judgment. Shipper's example is Every itself: the company uses agents for stable, repeatable work while keeping humans involved in customer service, writing, editing, engineering, and complex decisions. 4

Whittemore adds a broader definition of work through Paul Millerd's contribution. Caring for children, cooking, writing, maintaining relationships, and building a meaningful life all involve work, even when a paycheck is absent. That broader frame matters because employment statistics describe jobs, while people's experience includes many responsibilities that never appear in a job description. 1

The episode's organizational argument goes beyond putting a chatbot inside an existing workflow. Sumit Singh describes two possible uses for AI. Efficiency AI makes a familiar task faster or cheaper. Opportunity AI asks what a company can do because AI makes a new kind of work feasible. The first path gives a team a sensible place to begin. The second path requires experimentation because the answer cannot be copied from the old process. 1

Singh's point applies to product design as well as internal operations. A company that simply adds AI to an old workflow may preserve the old workflow's limits. A company that asks what becomes possible with an agent can redesign the sequence itself. Whittemore compares this change with the mobile shift that produced Uber, DoorDash, and Instacart: those companies used the phone's location and connectivity to create new services instead of reproducing a paper dispatch process on a screen. The episode presents the analogy as a way to think, rather than as evidence that AI will produce a known set of winners. 1

Noah Breyer, cofounder of Alephic, shifts the focus from throughput to direction. His argument is that agentic engineering's serious failure mode is a feature or product that works technically while moving away from the company's vision, values, or architecture. A software factory can optimize repeated production. A software company still has to decide what it should build and keep a team of humans and agents pointed at that decision. 1

Tom Critchlow gives the coordination problem a name: the company with the best clock may beat the company with the best model. Agents can act in seconds, teams meet weekly, finance plans quarterly, and leadership revisits strategy annually. Those different tempos leave each part of an organization with a different view of the present. Critchlow proposes a continuously updated record of goals, decisions, permissions, and constraints—a shared status layer for humans and agents. The episode also keeps the limit in view: synchronized information cannot decide what a team ought to value. 1

Karri Saarinen, cofounder and CEO of Linear, describes the related product-design problem. Traditional interfaces assume that a person navigates menus and watches each action. An agent may act first, leaving people to understand what happened and why. Saarinen argues that reliable AI products will need interfaces that give both humans and agents more structure, context, and visibility into the work. 5

The episode's final group of essays asks which kinds of work grow after automation. Nir Zicherman, CEO of Oboe, expects fewer roles built around concrete, verifiable tasks and more work involving ambiguity, open-ended creation, communication, organizational management, and oversight of many agents. His film-industry example separates the creative act from the surrounding logistics: AI may streamline financing, casting calls, and production management, while lower costs let human artists attempt projects that once looked technically or economically out of reach. 1

Joe Hudson calls the rising category wisdom work. His examples include emotional clarity, discernment, connection, and the ability to act when the answer cannot be verified in advance. Sari Azout, founder and CEO of Sublime, describes the same movement as attention being handed back to people. AI can automate a repetitive claim-processing task; the human choice is whether the recovered attention goes into more busywork, a new experiment, a relationship, or a decision that carries personal responsibility. 1

The project also gives human advantage a less polished name. Bethany Crystal argues that AI can make room for unusual interests and personal experiments because one person can now produce projects that once required a larger team. Emily Vernon argues that cheap, polished output will make a distinctive idea more valuable because a passable idea can be generated quickly. In both cases, the scarce contribution is a choice: what deserves to exist, what feels alive, and what someone is willing to stand behind. 1

For an individual or team, the practical test is to inspect the work beneath the job title. Which steps repeat cleanly? Which steps depend on current context? Where does someone define success? Who catches a plausible but wrong result? Which decisions carry consequences that an automated process cannot own? The answers reveal whether AI should speed up the existing workflow or create room for a different one.

Whittemore's episode is useful because it replaces a single employment forecast with a set of choices about work. Automate the repeatable layer. Spend the recovered attention on better questions. Build the coordination and interface systems that let agents operate alongside people. Keep human responsibility attached to the decisions that determine what a team, product, or life is for.

Listen to the full episode on Apple Podcasts. 1