Marcus Andrey Study
Ler em português Library

AI for UX · For leads · 5 min read

Guiding designers through AI adoption

Leading AI adoption in a design team is change management, not a tool rollout. How to see where each person is stuck (ADKAR), how to talk about fear honestly, a six-week learning path, the structures that keep it going, how to protect craft and growth, and what not to do.

Five seedlings, one for each stage of adopting AI (ADKAR): awareness, desire, knowledge, ability, reinforcement. Each person is at their own stage. Point at a pot and that plant grows a stage, while the others lean towards it. Illustration: Marcus Andrey, drawn with Hairline (MIT, Lucas Marques).

Buying licences is the easy part. The hard part is people: some designers are excited, some are quietly worried about their jobs, some tried a tool once, got a bad result and gave up. If you've ever led a move to a new design tool or a new design system, you already know the playbook. AI adds one more layer: it touches people's sense of what makes them valuable. This section is for anyone who leads designers, formally or not.

What you'll be able to do after reading:

  • Use ADKAR to see where each designer is stuck, and what helps at that stage
  • Open the conversation about fear, honestly
  • Run a six-week learning path with real tasks
  • Set up the structures that make adoption last
  • Protect craft and growth, especially for junior designers

1. See where each person is stuck: ADKAR

ADKAR is a model of individual change created by Jeff Hiatt at Prosci: people move through five stages, in order, and change stalls at the first one that's missing (Prosci). It works per person, not per team, which is exactly what AI adoption needs.

AAwareness

Do they know what AI can do for their own tasks?

Try: real examples from your team, not hype videos
DDesire

Do they want to use it?

Try: talk about fear openly; show it removes boring work, not judgement
KKnowledge

Do they know how?

Try: hands-on sessions, a playbook, pairing
AAbility

Can they do it on real work?

Try: short pilots on real tasks, office hours
RReinforcement

Does it stick?

Try: show and tell, recognition, update the definition of done
ADKAR, per person. Change stalls at the first missing stage, so help with that one first. Diagram: Marcus Andrey, CC BY 4.0.
Seedling trays on a greenhouse bench: some still bare soil, some with tiny sprouts, some with full leaves.
Same bench, same care, different stages: some trays have just been sown, others are ready to plant out. A team adopting AI looks like this, and each tray needs something different. Photo: TwHerbGrower, cropped, Wikimedia Commons, CC BY-SA 4.0.

2. The conversation about fear

Some designers will worry, often silently. As a lead:

  • Name it first. "A lot of designers are wondering what AI means for their role. Let's talk about it openly."
  • Be honest. Roles are changing: less production, more framing, specifying, reviewing and domain expertise. Don't promise that nothing will change.
  • Show the path. Link new skills to growth and recognition (see Career ladder).
  • Make space for sceptics. Their concerns about quality, ethics and data are valid and make the process better. Invite them to own the AI design review process.
  • Use One-on-one meetings for individual worries and retros for team-level ones. See Psychological safety.

3. A six-week learning path

  1. Week 1Guardrails

    Rules and approved tools, on a safe task.

  2. Week 2Writing for AI

    A spec and prompts for a real feature.

  3. Week 3Generate in the system

    Options with design system components.

  4. Week 4Code prototypes

    Pair with a developer.

  5. Week 5Research

    AI synthesis, checked by hand.

  6. Week 6Review and share

    Peer review, then show and tell.

A six-week path, as a cohort, on real but non-sensitive work. Diagram: Marcus Andrey, CC BY 4.0.

Run it as a cohort across product areas, a few hours a week, on real but non-sensitive work. It also connects designers who otherwise work alone in their product teams.

Week Hands-on task Output
1 Read your AI usage guidelines; try the approved tools on a non-sensitive task A personal list of tasks to try with AI
2 Write a design spec and prompts for a real feature A spec reviewed by a peer
3 Prototype in Figma AI or Claude Design with design system components Two or three options, self-reviewed
4 Pair with a developer: Generating design with Claude Code A working prototype with real components
5 Synthesise anonymised notes with an approved tool, then check by hand (AI-assisted synthesis, step by step) AI and human synthesis, compared
6 Review each other's work with the AI design review process; present at a show and tell One entry in the team playbook

4. Structures that keep it going

  • Champions: as a rule of thumb, one per three to five designers. They try new tools, write playbook entries and run office hours. Rotate the role. See AI champions network.
  • A playbook: one wiki page per use case: task, tool, prompt or spec template, review notes, time saved, pitfalls.
  • Time to experiment: a protected slice of time in the first quarter, for structured experiments, not side-of-desk tinkering.
  • Show and tell: fifteen minutes in the monthly team meeting. Failures welcome.
  • Shared prompts, specs and skills: versioned, in one place. See Prompt patterns for designers and UX agents and skills: a guide for designers.

5. Protect craft and growth

  • Junior designers: alternate with AI and without AI on core skills, and always ask "why is this right?". Critique AI output together.
  • Senior designers: move them towards framing problems, writing specs, system design and mentoring.
  • Everyone: time with the people who use your product matters more than ever. It's the knowledge AI doesn't have. See Contextual inquiry.

6. Common mistakes

  • Usage quotas, or tracking individual AI usage. It measures compliance, not value, and it destroys trust. In NN/g's 2026 survey of UX practitioners, forced adoption was one of the recurring complaints (NN/g).
  • Announcing "AI will make us 30% faster" before there's evidence. In one careful study, developers expected AI to make them 24% faster and were measured 19% slower (METR, 2025). See Measuring AI features.
  • Letting AI-generated screens skip critique because "it's just a draft".
  • Rolling out an unapproved tool because one designer liked it. See Choosing and piloting AI tools.
  • Assuming everyone learns the same way, at the same pace.

7. Vocabulary

Term Meaning
ADKAR Awareness, Desire, Knowledge, Ability, Reinforcement: five stages of individual change
Change management Helping people move from the old way of working to the new one
Champion A designer who helps others adopt a tool and collects what works
Cohort A group that learns together, on the same schedule
Playbook A shared collection of proven ways to do a task

8. Try it (no code needed)

  1. List the designers you work with (only for yourself; never share it).
  2. For each, guess their ADKAR stage, and the one thing that would help them move.
  3. Check your guesses in a one-on-one: "What would make AI more useful for you right now?"
  4. Pick one structure from section 4 that you can start this month.

Reflection questions: Which stage is your team stuck at most? Where do your own doubts about AI sit, and how would you want a lead to talk to you about them?


Checklist

Learn more

Books, courses and videos to go deeper

Book links go to Amazon. As an Amazon Associate I earn from qualifying purchases, at no extra cost to you.

Choosing and piloting AI tools · UX agents and skills: a guide for designers · Building an AI agent roadmap for your UX team · AI adoption ladder · Kotter 8 steps · Situational leadership · Individual development plan · Module 12: AI and the UX Lead

Saved only on this device.

↑ ↓ to move · Enter to open · Esc to close