The design process is dead. Here’s what’s replacing it. | Jenny Wen (head of design at Claude)

Summary of The design process is dead. Here’s what’s replacing it. | Jenny Wen (head of design at Claude)

by Lenny Rachitsky

1h 17m•March 1, 2026
Listen to Episode

Overview of The design process is dead. Here’s what’s replacing it. (Lenny Rachitsky with Jenny Nguyen)

Jenny Nguyen (lead of design for Claude / Anthropic, formerly Figma, Dropbox, Square, Shopify) joins Lenny to explain how AI-driven engineering workflows are forcing a major shift in how designers work. The old multi-month/quarter design process — heavy on research, high-fidelity mocks and long-term vision decks — is giving way to two fast-moving modes: (1) design that supports rapid execution (pairing with engineers, last-mile polish, in-code prototyping) and (2) short-horizon vision work (3–6 months, prototype-first). Jenny shares concrete changes to daily work, tooling, hiring, leadership practices, and product strategy (illustrated via Anthropic’s Claude and Co-Work launches).

Major shifts in the design process

  • The old “trust the process” model (long discovery → diverge/converge → polished mocks) is dying — accelerated by AI that lets engineers prototype and ship faster.
  • Work splits into two primary modes:
    • Execution-support (consulting, pairing, code polish, quick prototypes).
    • Short-horizon directional vision (3–6 months; rapid prototypes rather than long decks).
  • Practical change in time allocation (example from Jenny):
    • A few years ago: ~60–70% mocking/prototyping, ~20% engineering collaboration.
    • Today: ~30–40% mocking/prototyping, ~30–40% jamming/pairing with engineers, plus some implementation.

Two types of design work (how to think about them)

  1. Supporting implementation

    • Let engineers “cook” and ship quickly; designers unblock rather than gatekeep.
    • Provide principles, components, code snippets, and pair on last-mile polish.
    • Equip engineers with design-system code examples so future work requires less hand-holding.

  2. Direction & vision

    • Short, tactical visions and prototypes that point the org toward coherent product decisions.
    • Focus on surfacing the right experiments and use cases rather than long-term rigid roadmaps.

Day-to-day at Anthropic (what Jenny actually does)

  • Keep up with internal research and prototypes across many teams (big information flow).
  • Prototype in code and in Figma; often pairs with engineers on implementation and PR-level polish.
  • Reads internal signals to spot illegible but high-potential ideas; extracts UX/formatting patterns.
  • Labels many releases as research previews and relies on fast iteration and visible response to feedback to maintain user trust.

Tools / AI stack Jenny uses

  • Anthropic/Claude family: Claude chat, Claude Co-Work (for longer-running tasks & multi-step workflows), and a code-assist tool (referred to as Claude Code).
  • VS Code (especially when pairing on front-end changes).
  • Figma — still important for divergent exploration, multiple directions, and micro visual/interaction decisions.
  • Other: Slack integrations, lightweight editors (e.g., iA Writer), and personal apps like Retro for reflection.

Key idea: designers now need a mixed toolkit — canvases for exploration (Figma) plus in-code prototyping and agent-driven tooling.

Product strategy & release philosophy

  • Ship early where the value justifies it, mark things as research preview, and iterate visibly.
  • Build trust through speed + responsiveness: show users you’re listening and improving quickly.
  • Use prototypes with real models (not only mocked states) for AI-driven features — models are non-deterministic and usage reveals real cases.

Hiring: who succeeds in this new era?

Jenny’s three archetypes she’s hiring for now:

  1. Strong generalists (block-shaped / broad, with multiple 80th-percentile skills) — adaptable across PM/engineering/design tasks.
  2. Deep specialists (T-shaped with a very deep tip: e.g., visual design, interaction, or technical design) — to differentiate products.
  3. “Cracked new grad” — early-career people who are unusually curious, resilient, humble, and fast learners (less baggage, rapid adoption of new tools/processes).

Practical advice for candidates:

  • Build things — ship projects, use modern tools, show curiosity.
  • Show evidence of learning and applied work (not only theory).

Management, leadership & team culture

  • Managers should consider rotating into IC roles (or at least do hands-on work) to empathize with changed workflows.
  • Low-leverage tasks done by senior leaders can actually be high-leverage:
    • Dogfooding, testing, filing bug PRs, and nitty-gritty work create credibility, context, and morale.
  • Psychological safety + high standards:
    • Encourage candid feedback (Jenny describes playful “roasting” among teammates as a sign of trust).
    • Balance warmth and direct challenge (similar to “radical candor”).
  • Managers now often need to combine people management with direct craft direction.

Frameworks & mental models

  • Legibility framework (Evan Tana): considers whether founders and ideas are legible or illegible. Designers act like internal VCs: spot “illegible” ideas (frontier prototypes with energy but unclear form) and help make them legible via storytelling, UX, and product form factors.
  • Use Slack/internal prototype feeds to find emergent concepts worth investing design attention in.

UI / interaction trends

  • Chat UIs are not going away — they provide a general-purpose interface that scales with model intelligence and allow rich interaction.
  • Expect hybrid surfaces: chat + interactive widgets (clickable components, to-do lists, previews). Jenny expects models to generate/assemble UIs more often.
  • Co-Work approach: “Claude with hands” — shows tasks Claude is working on, exposes plans/to-dos, and makes the model’s work tangible and actionable.

Notable quotes (high-signal soundbites)

  • “Don’t trust the design process” — the canonical long-process is dying.
  • “You as a designer actually do not have the time to make these beautiful mocks anymore.”
  • “We’re building trust through speed” — ship early as research preview, iterate visibly.
  • “Someone still needs to be accountable for the decision.” — human judgment remains crucial.

Practical takeaways & recommended actions

For designers:

  • Get comfortable pairing with engineers, prototyping in code, and doing last-mile polish.
  • Maintain Figma for divergent exploration, but learn code-based tools/agents that can ship product-ready experiments.
  • Choose one of three career shapes: strong generalist, deep specialist, or “cracked new grad” (learners who ship).
  • Build a portfolio of actual shipped projects and experiments (not just static mockups).

For managers:

  • Do occasional IC rotations or hands-on work to stay fluent with tools and team realities.
  • Do visible, nitty-gritty tasks (dogfooding, bug PRs) to build credibility and culture.
  • Cultivate psychological safety while holding high standards — candid culture aids rapid iteration.

For product teams:

  • Prefer short horizons and prototype-first visions for AI features; rely on rapid real-user feedback.
  • Label early launches as research previews and commit to visible iteration to preserve trust.

Where to follow / hiring note

  • Jenny online: X/Twitter — @jenny_wen (she encourages sending product feedback for Claude/Co-Work).
  • Anthropic / Claude design team is hiring and actively experimenting — they look for people excited by frontier tools and comfortable with rapid change.

If you want to understand the current and near-future shape of product design in AI-driven organizations, this episode is a concise field guide: expect more in-code prototyping, tighter designer-engineer partnership, shorter product horizons, and a premium on adaptability and judgment.