Stratified Design Work
Split your time between supporting engineers' execution and setting a 3-6 month vision, not polishing mocks.
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 3
- Confidence
- 90%
With AI letting engineers ship features constantly, the classic linear design process (research, diverge/converge, then hand off beautiful mocks) is obsolete. Jenny Wen argues design work now splits into two distinct modes: (1) supporting implementation and execution alongside fast-shipping engineers, and (2) creating a near-term vision that points everyone the same direction. The old 60-70% spent mocking drops to 30-40%, with the rest going to jamming with engineers and hands-on implementation.
Origin
Framed by Jenny Wen (head of design at Claude, ex-Figma director) in a September Berlin conference talk she called 'don't trust the design process,' and updated on Lenny's Podcast as the shift accelerated after Opus and Claude Code adoption.
Core principles
- 01The design process 'treated as gospel' is dead because engineering, not design, changed first
- 02Designers no longer have time to make beautiful mocks for every idea
- 03Don't block fast-shipping engineers; let them build the scrappy version and try it
- 04A shared vision makes many parallel efforts cohere and is more efficient than random directions
- 05With non-deterministic AI models you must discover use cases by watching real usage, not mocking every state
How to run it
- 1
Let engineers execute unblocked
When anyone floats an idea, let an engineer build a scrappy version and try it rather than gating on a designer's polished mock. Engineers are still better at implementing than designers.
Pro tip Shift some of your own time into last-mile polish and prototyping in real code so you participate in execution instead of guarding it.
Watch out Don't try to mock up every state of an AI feature; non-deterministic models mean states and use cases must be discovered from real usage.
- 2
Set a 3-6 month vision, not a 2-10 year deck
Because the technology changes too fast to know what happens in two years, replace the long-horizon vision deck with a near-term (3-6 month) direction, often expressed as a prototype rather than a beautifully story-told deck.
Pro tip Deliberately block time for this; execution mode crowds it out, so schedule vision work explicitly.
- 3
Rebalance your time pie
Move from ~60-70% mocking/prototyping to roughly 30-40% mocking, 30-40% jamming and pairing directly with engineers, and a growing slice of actual implementation and shipping.
Pro tip Use an IDE (not just an agent CLI) when you're tweaking frontend polish so you can see the code and talk to Claude at the same time.
In the wild
After a long stretch of pure execution mode on Co-work, Jenny set aside time to think about what the next three months could look like given where the market and models are, then visualized it to point the whole team the same way.
→ The team gets a shared near-term direction instead of drifting across the infinite feature directions that many parallel agents could produce.
Common mistakes
Clinging to the old linear process
Insisting on full discovery, diverge/converge, and polished handoff mocks now blocks fast engineering and wastes the designer's scarce time; the process was already dying before AI.
Trying to mock your way to certainty with AI features
You can't theorize or make a clickable prototype for non-deterministic models; you have to ship on the real model and discover use cases from how people actually use it.
Is it for you?
Best for
Product designers and design leaders at fast-shipping, AI-native product teams where engineers deploy features many times a day
Not ideal for
Regulated or slow-release environments where quality gates precede launch, or teams that ship rarely and can still afford long discovery cycles
From the transcript
“that is that's basically dead”
“there are basically two types of design work”
“you as a designer actually like do not have the time to make these beautiful mocks anymore”
“you're better off not blocking that letting them cook”
“it usually becomes a vision that's like 3 to 6 months out”
From the episode
The design process is dead. Here’s what’s replacing it.
Jenny Wen (head of design at Claude)