AI is rebuilding the product development workflow from the ground up, and the roles around it are being restructured along with it. The org chart we've operated with for two decades — product manager, designer, front-end/back-end/data/mobile engineer, functional and automation QA, plus the supporting cast of TPMs, product analysts, user researchers, content teams, platform and DevOps teams — was designed around the constraint that humans do the execution. As AI agents absorb more of that execution, the constraint changes, and thought leaders at the frontier are converging on new taxonomies for what humans actually do. Three frameworks stand out.
Boris Cherny's five archetypes
Boris Cherny, the creator of Claude Code at Anthropic, observes that as engineering, product, design, and data science melt into a single kind of role, what remains are five archetypes, drawn from watching his own team:
- Prototyper — comes up with brand-new ideas and churns out many of them, most of which never ship.
- Builder — quickly turns a prototype into production-grade product and infrastructure.
- Sweeper — cleans up the UI, simplifies the code and system, unships what isn't pulling its weight, and optimises performance.
- Grower — takes a built product and iterates on it toward product-market fit.
- Maintainer — owns a mature system and keeps it secure, reliable, fast, and efficient as it scales.
Two details make this framework interesting. First, most people span two archetypes, sometimes three — these are modes of contributing value, not job titles. At Anthropic, some designers are Prototypers and some engineers are Maintainers; the badge title predicts almost nothing. Second, the right mix changes with the product lifecycle: an early product needs Prototypers and Builders, a mature one needs Sweepers, Growers, and Maintainers.
Zeb Evans' three-role "100x org"
Zeb Evans, CEO of ClickUp, restructured his entire company around AI agents in May 2026 (roughly 3,000 internal agents, a 3:1 agent-to-employee ratio) and defined three human roles:
- Builders — 10x engineers and 10x product managers. His argument: the best engineers no longer write code, they direct agents that write code. Judgment and review are the skills that matter.
- Agent managers (system managers) — people who automate their own workflows and then own the resulting agent systems: monitoring the fleet, managing costs, handling escalations.
- Front-liners — customer-facing people who spend nearly all their time with customers. In a world saturated with AI-generated communication, genuine human contact becomes the one bottleneck worth protecting.
Evans also notes that product management and design are merging: designers with customer instinct become PM-like, and PMs with UX intuition become designer-like, because agents have removed the user-research bottleneck between them.
The four-role governance view
A third framing, via Foundation Capital, looks at it through the lens of accountability rather than production, identifying four broad human roles:
- Chief accountability officer — the human who owns outcomes when agents do the work.
- System architect — designs how humans, agents, and infrastructure fit together.
- Relationship expert — owns the human relationships that AI cannot replace.
- Validators — verify, review, and quality-assure agent output before it ships.
What the frameworks agree on
The terminology differs, but three threads run through all of them. Titles are dissolving into modes of value — what you're called matters less than which phase of creation you're good at. Human judgment moves up the stack — from executing work to directing, reviewing, and taking accountability for agent output. And human relationships become premium — as AI saturates every communication channel, the customer-facing human becomes more valuable, not less.
For anyone in a traditional product role today, the practical question is not "which of my tasks will AI take?" but "which of these archetypes am I, and which does my product's current lifecycle stage need?"