The 5 Roles in AI-First Companies
How engineering, product, design, and data science are melting into a new kind of role — and what healthy teams actually need at each stage
← Part of the AI knowledge base · Productivity knowledge base
Source: Post by Boris Cherny (@bcherny), creator of Claude Code — Read the original post on X
The Old Job Titles Are Blurring
Engineering, product, design, data science, and the rest are melting into something new. The domain-specific titles we inherited from the last decade — "frontend engineer," "product manager," "UX designer" — describe what someone works on, not how they contribute to a product's lifecycle.
When Boris Cherny looked at the Claude Code team at Anthropic, he saw five archetypes that cut across traditional job functions. These aren't departments. They're modes of contribution — and the best people often span two or three at once.
The Five Archetypes
1. Prototyper
Comes up with brand-new ideas and churns out many of them. Most prototypes don't ship — and that's the point. Prototypers explore the possibility space fast, testing what might work before anyone commits to building it for real.
2. Builder
Quickly turns a prototype or idea into production-grade product and infrastructure. Where the Prototyper asks "what if?", the Builder answers "let's ship it." Speed and quality at the same time.
3. Sweeper
Cleans up the UI, simplifies the code and system, unships what doesn't belong, and optimizes performance. Sweepers make things feel finished — the craft layer that separates a demo from a product people trust.
4. Grower
Takes a product that has been built and iterates on it to improve product-market fit. Growers watch how real users behave, run experiments, and push the product toward the version of itself that people actually want.
5. Maintainer
Owns a mature system to keep it secure, reliable, fast, and efficient as it scales. Maintainers are the reason a product that found PMF doesn't collapse under its own success.
These Roles Cross Job Functions
This is not an engineer thing or a designer thing. Across Anthropic, Cherny noticed that some designers match category 1 (Prototyper), some match 2 (Builder), some match 3 (Sweeper) — and the same is true for engineers, PMs, and data scientists.
The archetype describes how you work, not what your LinkedIn title says. A designer who ships fast and thinks in systems might be a Builder. An engineer obsessed with polish and simplification might be a Sweeper. A PM who lives in experimentation loops might be a Grower.
Many people span across two roles, and sometimes three. That's normal — and often a sign of a strong contributor.
What Your Team Needs Depends on the Product Stage
A healthy team needs a mix of these archetypes, but the right mix shifts as the product matures.
| Product Stage | Primary Archetypes Needed | Why |
|---|---|---|
| New & Pre-PMF | Prototyper + Builder + Sweeper (1 + 2 + 3) | You need ideas, fast execution, and enough polish to learn from real usage — not a perfect system. |
| Growing & Found PMF | Builder + Sweeper + Grower, plus some Maintainer (2 + 3 + 4 + some 5) | The product works. Now you iterate toward stronger fit, clean up what shipped fast, and start hardening what scales. |
| Strong PMF & Mature | Sweeper + Grower + Maintainer, plus some Builder (3 + 4 + 5 + some 2) | Growth and reliability dominate. New features still matter, but the core job is keeping and improving what already works. |
What This Means for Hiring and Team Design
- Hire for archetype, not just function. A "senior engineer" who only prototypes and never ships is a different bet than one who builds and sweeps. Know which gap you're filling.
- Audit your team's shape. If you're pre-PMF but everyone is a Maintainer, you'll move too slowly. If you're mature but everyone is a Prototyper, you'll never stabilize.
- Let people wear multiple hats. Two or three archetypes in one person is a feature, not a bug — especially in small, AI-augmented teams where leverage comes from range.
- AI changes the ratios. When one person with Claude Code can prototype, build, and sweep in a single afternoon, the bottleneck shifts from "can we ship?" to "do we have the right mix of judgment across the lifecycle?"
The Bigger Picture
Maybe product roles of the future will look more like these five archetypes — and less like the domain-specific silos of today. Engineering, product, design, and data science aren't disappearing. They're dissolving into a shared set of contribution modes that map to what a product actually needs at each stage of its life.
For AI-first companies, where small teams move at extraordinary speed, getting this mix right may matter more than any org chart.
Related Roles, Frameworks & Experiments
Loops, Not Prompts — Boris Cherny
From the same creator of Claude Code: how top teams shifted from prompting to writing autonomous loops — and why generalist builders are winning.
Read the deep dive →AI Engineer vs AI Operator
Another lens on how AI is reshaping job functions — the classic technical path vs. the emerging operator who orchestrates agents in real workflows.
Full comparison →The Distribution Engineer
The emerging role for builders who treat distribution like engineering — and why hiring for archetype matters more than title.
Read the deep dive →AI Chief of Staff
One persistent orchestration layer that delegates across projects — how small AI-augmented teams multiply output without adding headcount.
Read the deep dive →AI Roles & Character Definitions
On the prompting side: giving AI precise role definitions dramatically improves output quality — a complement to team archetypes.
Read the deep dive →Mike's CEO Operating System
A markdown-based productivity OS for running multiple companies with AI agents — team design in practice.
Read the OS →Browse more in the AI category and Productivity category.
Idea and framework from Boris Cherny (@bcherny), creator of Claude Code at Anthropic.
View the original post on X →