§ Playbook · Design
The AI playbook for Design
AI is not killing design - it is splitting it. Production design shrinks while product design grows, and the design system becomes either your best asset or your most bypassed one. This playbook covers the split, the roles, and the 90 days that decide which side you land on.
Free · no form, no follow-up sequence
Figma's own research says 91% of designers now use AI weekly - and 32% trust what it produces. That gap is not a bug in the tools. It is the job description of every designer you keep.
91% / 32%
designers using AI weekly vs designers who trust its output.
Figma AI reports, 2025-26
11th
graphic design's rank among fastest-declining jobs - while UX ranks 8th fastest-growing.
WEF Future of Jobs, 2025
10 of 12
technical writers laid off at Canva, months after an AI mandate.
ACS Information Age, 2025
§ Part one
What actually changes.
The World Economic Forum's jobs data tells the story in one line: graphic design is now the 11th fastest-declining job category, and UI/UX design is the 8th fastest-growing. Same report, same year. AI is not coming for design - it is coming for production, and it is promoting everything else.
The tools are genuinely quick. A working prototype in fifteen minutes is real. So is the catch: practitioners describe the output as a high-fidelity wireframe with the wrong design decisions embedded - not accessible, not semantic, not on your system. Fast to first draft; expensive to done, unless someone with taste is steering.
01
Prototyping collapses to minutes
Figma Make, v0, Lovable and friends turn an idea into a clickable thing before the meeting ends. Wonderful for direction-setting and killing bad ideas early. Dangerous the moment anyone mistakes it for shipped work.
02
Production design mostly stops being a role
Resizing, versioning, asset adaptation, UI copy variants - the volume work that filled junior portfolios is machine work now. The WEF decline number is this, measured.
03
Research synthesis genuinely speeds up
Single-study qualitative analysis is the strongest evidenced use case. AI degrades on multi-study synthesis, where the nuance lives - so it does the transcription and clustering, and your researchers do the meaning.
04
Non-designers now generate UI
PMs and engineers can produce plausible screens without opening Figma or touching your design system. That is not a hypothetical governance question. It is happening in your organisation this week.
05
The design system becomes AI infrastructure
Tokens and component docs structured for machine consumption make AI output dramatically better. A design system that is not AI-legible is a very well-documented museum.
06
Taste becomes the deliverable
When anyone can make something plausible, the value moves to knowing which plausible thing is wrong. Figma's 32%-trust number is your job security, if you organise around it.
§ Part two - the honest bit
What it costs you.
Total demand for designers looks stable - 82% of hiring managers say need has held or grown - but composition is shifting hard against juniors and pure-production roles, and 45% of designers already feel it. Both things are true: the function survives, some of the jobs in it do not.
Canva is the cautionary tale nobody in design wants to discuss: it told staff AI would not cause layoffs, mandated AI adoption, and then let ten of its twelve technical writers go. The lesson is not that Canva is wicked. It is that reassurances outrun plans, and your people know it.
01
Junior roles thin out first
The production tasks that trained juniors are automating away. If you still want senior designers in five years - and you will - you must now build the apprenticeship on purpose: paired critique, real problems, judgement reps.
02
The design system paradox lands on you
Your system is simultaneously more valuable (it is the only guardrail that makes AI output usable) and more bypassable (AI generates plausible UI without it). Make it AI-legible or watch the investment strand.
03
Design debt now compounds at machine speed
AI tools do not know which modal variant you standardised last sprint. Off-system output that looks on-brand is the most expensive kind, because nobody flags it until it is everywhere.
04
Craft roles adjacent to design take the hit
Technical writing at Canva. Photography and modelling at H&M, replaced by digital twins. If your function commissions craft, the commissioning is changing whether you plan for it or not.
05
The Figma line item becomes a question
When PMs prototype in vibe-coding tools and engineers generate UI in their editor, seat counts and tool boundaries stop being settled. Do not assume the current stack survives its next renewal unexamined.
06
Quality failures are now design failures
The Coca-Cola backlash, the uncanny faces, the plastic-looking animals - every public AI embarrassment was, at root, a design quality decision someone did not make. That accountability is moving to your desk. Take it in writing.
§ Part three
What good looks like.
A good design function ends up smaller in production, stronger in judgement, and central to how the whole organisation uses AI - because design is where taste, systems thinking and user evidence already live. The teams that thrive stop defending tasks and start owning quality.
01
Designers direct, machines produce
Seniors set direction, prototype with AI at speed, and gate what ships. The 15-minute prototype is used to think faster, not to ship faster than thinking allows.
02
An AI-legible design system
Tokens, components and usage rules structured so the tools consume them - which means AI output lands on-system by default, and the bypass problem shrinks to an exception process.
03
Research with humans on the meaning
AI handles transcripts, clustering and first-pass synthesis. Researchers spend their recovered hours on the multi-study, contradictory, actually-hard questions - and on being in the room where decisions happen.
04
A designed apprenticeship
Juniors learn judgement through critique and real product problems, not through the production work that used to double as training. Fewer hires, better formed, with a visible path to senior.
The change, across all five pillars
People
Production thins, direction grows - and the junior path is rebuilt around critique and judgement reps.
Processes
A named quality gate between generated work and shipped work, wired into how prototypes become product.
Data
Research synthesis accelerated by machine, with humans kept firmly on the meaning.
Platforms
A design system made AI-legible, so generated output lands on-system by default instead of around it.
Partners
Commissioned craft - writing, photography, illustration - renegotiated deliberately before it drifts.
§ Part four
The first 90 days.
The goal is not to adopt tools - your team did that without asking. The goal is to put quality, systems and roles back in charge of what they adopted.
Ⅰ
Weeks 1-2
Audit the adoption you already have
Find out what tools are actually in use, by whom, on what work, and what has shipped off-system. No blame - you need the real map, and you will not get it twice.
Ⅱ
Weeks 3-6
Make the system AI-legible
Structure tokens, components and rules for machine consumption. Pick one product surface and prove that AI output can land on-system. This is the highest-leverage design work of the year.
Ⅲ
Weeks 7-10
Put a quality gate in front of generation
Define what 'should a machine have made this?' means for your brand, name an owner, and wire the gate into how prototypes become production. Measure rework before and after.
Ⅳ
Weeks 11-13
Redesign the roles, out loud
Publish the honest version: production shrinks, direction and systems grow, and here is the junior path we are building. Your designers have read the WEF report too.
The day-90 scorecard · what you should be able to show
§ Part five
The traps, so you can skip them.
Trap 01
Banning the tools
Your team is already using them; a ban just moves the usage to personal accounts and removes your visibility. Govern what exists rather than legislating against the tide.
Trap 02
Trusting the demo
The prototype looks finished because looking finished is what these tools are best at. Accessible, semantic, on-system and correct are separate properties, and none come free.
Trap 03
Defending tasks instead of quality
Fighting to keep production work in human hands is a losing position held for honourable reasons. The winning position is owning what good looks like, wherever the pixels come from.
Trap 04
Letting the system rot
Every week the design system stays machine-illegible, more off-system UI ships around it. The asset does not strand on a decision day; it strands by a thousand bypasses.
Trap 05
Starving the pipeline quietly
Not hiring juniors feels like a savings line until your senior bench retires. If the apprenticeship automated away, build a new one - deliberately, and in the budget.
Trap 06
Promising no job changes
Canva promised. Ask its technical writers. An honest account of the split - production down, judgement up - is uncomfortable once; a broken reassurance is expensive forever.
§ Before you go
Six questions worth an honest hour.
Score yourself before someone with a rate card does. If more than two answers are uncomfortable, that is not a crisis - it is an agenda.
The self-assessment
§ Start with the workshop
Book your
Catalyst Workshop.
Bring the strategic change you have been circling. In 1-2 executive days, we will turn it into the first move, the case for it, and the mobilisation plan.