AI job titles are becoming work modes

The useful idea in Nathaniel Whittemore's July 5 episode is not that every company should rename its job titles. It is that AI agents make work more decomposable: people can generate prototypes, harden them, prune them, grow them, maintain them, select among them, and manage the risks around them faster than old functional charts were built to handle. The episode uses Boris Cherny's Claude Code team archetypes as a starting point, then pushes the frame beyond engineering into sales, marketing, finance, HR, and operations 1.

Whittemore explicitly sets the episode apart from the usual displacement debate. He is not asking which industries lose headcount, or which brand-new job titles appear. His frame is narrower and more operational: when people can hand execution to agents, what remains inside the human role? The answer is not one new universal title. It is a set of recurring modes of work 1.

That distinction matters because it avoids the weakest version of future-of-work analysis. Saying "AI will create prompt engineers" is too shallow. Saying "AI will automate jobs" is too coarse. The better question is what parts of a role become cheap enough to repeat, and what judgment becomes more expensive because repetition is cheap.

In this episode, cheap execution makes two things scarce: selection and stewardship. Selection means choosing which of many possible artifacts should receive real attention. Stewardship means keeping speed from turning into unmanaged risk.

The source framework comes from Boris Cherny, who wrote that as engineering, product, design, and data science melt into a new kind of role, the Claude Code team appears to contain five archetypes: prototyper, builder, sweeper, grower, and maintainer 2.

Cherny's distinction is useful because it cuts across job functions. The prototyper produces many ideas, most of which do not ship. The builder turns a promising idea into production-grade product or infrastructure. The sweeper removes complexity, unships what should not exist, cleans up interfaces, and improves performance. The grower iterates toward product-market fit. The maintainer owns a mature system so it stays secure, reliable, fast, and efficient as it scales 2.

Whittemore reads these less as job descriptions than as temperaments. One person may move through several modes over a project lifecycle, but organizations still need to know which mode they are underweight in. A pre-product-market-fit team needs more prototyping, building, and sweeping. A mature product needs more sweeping, growing, and maintaining 2.

The practical point is easy to miss: agents do not erase the lifecycle of work. They compress parts of it. A prototype can appear faster, but someone still has to decide whether it deserves to become a system. A feature can be implemented faster, but someone still has to simplify, secure, document, and operate it.

The episode's strongest move is adding roles that face the market, the organization, and the risk environment around the work. Whittemore argues that Cherny's five modes mostly face the artifact being built. They describe how a team moves from idea to system. What they miss is the signal loop around that system 1.

His added roles start with the editor. If prototypers can create a dozen plausible directions in a week, the editor chooses which ones deserve more resources. That role may use metrics, taste, customer evidence, or strategic judgment. The common function is focus.

The scout gathers signal from the outside world: customers, competitors, culture, workflow pain, and weak signals that could become new product or process ideas. Whittemore notes that parts of scouting may themselves become agentic because agents are well suited to collecting and distilling large information streams 1.

The evangelist translates an internal view of the world into a market-facing one. That is not just marketing copy. It is the work of making customers, users, employees, or a community understand why the new thing matters.

The orchestrator coordinates across archetypes and teams, especially when output increases. Whittemore separates this from the conductor, who is more directly responsible for managing agent fleets and making their outputs coherent with one another. The line between those two may blur, but the difference is useful: one coordinates organizational work, the other coordinates synthetic workers.

The risk steward may be the least glamorous role, but it is also the one that changes most under agentic speed. In slower organizations, risk often appears as a gate. In faster organizations, Whittemore argues, the risk steward becomes forward-looking: the person who anticipates what could derail the work before momentum carries the organization into trouble 1.

The episode becomes more useful when it leaves product teams. In sales, a prototyper might test pitches, segments, and offers. A builder turns the working version into a repeatable playbook. A sweeper cuts dead scripts, poor-fit segments, and low-value activity. A grower watches deal velocity and expansion. A maintainer keeps the sales system disciplined quarter after quarter 1.

Marketing maps even more cleanly. Scouts read the audience and competitors. Prototypers test narratives, campaigns, and channels. Editors decide which angle fits the brand. Builders turn those sparks into campaign machinery. Sweepers prune weak messaging and channel bloat. Growers optimize conversion and cost. Maintainers preserve brand and lifecycle systems 1.

This is where the episode's thesis becomes sharper: agentic work does not merely make existing departments faster. It pushes product-like behavior into functions that used to depend on tickets, handoffs, and quarterly roadmaps. A finance or HR operator may not become a software engineer, but they may become the person who builds a narrow internal tool for a recurring exception rather than waiting for a centralized team to prioritize it.

Whittemore's most concise claim comes near the end: when making gets cheap enough, every function starts to grow a maker 1. That is the line that gives the episode its real weight.

The maker is not simply the person who can use a coding agent. It is the person inside a function who understands the work deeply enough to prototype a better way to do it, then pull the right builder, editor, sweeper, maintainer, scout, orchestrator, or risk steward into the loop.

For practitioners, the near-term career advice is straightforward: learn where your function has repeated friction, then become capable of turning that friction into a working artifact. The artifact does not have to be a SaaS replacement. It can be a small tool, an agent workflow, a process wrapper, or a prototype that forces a better conversation.

For managers, the harder task is diagnostic. Do not ask only whether the team has enough engineers, PMs, designers, or analysts. Ask which modes of work are missing. Too many prototypers and no editors means noise. Too many builders and no sweepers means compounding complexity. Too many growers and no maintainers means brittle scale. Too much speed and no risk steward means the system eventually stops itself.

That is why this episode is worth reading as organizational design, not labor-market forecasting. It turns the AI jobs conversation away from "which title survives?" and toward a more useful question: once execution gets cheaper, which human work becomes the bottleneck?