Marcus Andrey Study
Ler em português Library

Module 12 of 12 · 12 min read

AI and the UX Lead: Guiding a Design Team Through AI

What AI changes in UX work, the principles and minimum governance a lead sets, how to guide each designer through adoption, and a 90-day AI plan for your team.

Why it matters

In 2026, every UX team is working out how AI changes its work, and most are doing it chaotically: individuals experimenting alone, unclear rules about data, AI-generated screens that bypass the design system, juniors worried about their future, and managers asking "can't AI just do this?"

A UX lead, especially one responsible for internal or enterprise applications, has to bring direction, safety and quality to that chaos. It's the same kind of change as any tooling migration (see Kotter 8 steps): a shift in tools that is really a change in how people work together. Your design-system knowledge, your sense of quality and your awareness of data-protection constraints are exactly the strengths this needs.

Common pitfalls for new leads:

  • Banning or ignoring AI. People use it anyway, just without guardrails.
  • Mandating it. Usage quotas and "everyone must use tool X" create resentment and shallow use.
  • Leading by being the best prompter. You end up doing the work instead of building the team's judgement.
  • Forgetting the juniors. They lose the struggle that builds their skill.

The lead's job isn't to be the best AI user on the team. It's to make sure the team uses AI well: on the right tasks, with the right data, at the right quality, and in a way that helps designers grow instead of being replaced.

The AI deep-dive path. Read this module first, then go step by step. Each note has ← → links.

Step Notes You'll be able to…
① Understand the tools AI tools landscape for designers → Figma AI → Claude Design pick the right tool for the fidelity you need, and check what's enabled in your organisation
② Work with AI yourself Prompt patterns for designers → Design specs for AI → Generating design with Claude Code brief AI well and turn a spec into an on-system prototype
③ Keep quality and safety AI design review process → AI governance and data protection review AI output against the bar, and stay within GDPR, the AI Act and works-council rules
④ Lead people through it Guiding designers through AI adoption → AI in UX research bring sceptical and anxious designers along, and use AI in research responsibly
⑤ Design AI into products Agentic design design AI features experts trust: suggestions, confidence, control, recovery

Worked example: Building an AI agent roadmap for your UX team and UX agents and skills: a guide for designers show how a UX team can build its own AI agents and skills step by step. Use them as a model for your own first pilots.


1. What AI changes in UX work

Activity Where AI helps today Where human judgement stays Main risk
Research planning Drafting discussion guides, screeners, survey questions Choosing what to learn and why Generic questions that miss domain nuance
Synthesis Clustering notes, first-pass themes, transcript summaries Interpreting meaning, weighing evidence, spotting what's missing Confident but shallow insights; hallucinated quotes
Ideation Many variations fast, analogies, edge cases Choosing direction, taste, strategy fit Sameness: everyone gets the same "average" ideas
Specs & documentation Drafting specs, acceptance criteria, annotations, release notes Deciding behaviour, trade-offs, edge-case policy Specs that look complete but aren't true
UI generation & prototyping Prompt-to-prototype, code prototypes with real components, layout options Information architecture, interaction model, fit to expert workflows Off-system UI, inaccessible output, "demo-ware"
Handoff Design-to-code via MCP and design-system mappings, generated annotations Defining what "done" means Code that matches pixels but not behaviour
QA & review Accessibility checks, consistency audits, visual diffing Final quality call False sense of safety from automated checks
Communication Summaries, decks, stakeholder updates The message, the narrative, the relationship Loss of voice and credibility

Pattern: AI compresses production (making artefacts) and expands exploration (seeing more options). NN/g's 2026 survey found practitioners work faster but have mixed feelings (AI in UX survey), and it argues critique is now the core design skill (article). It does not remove the need for judgement: framing the problem, choosing, and owning the outcome. For a lead, this means:

Shift team time from making to framing, deciding and reviewing, and raise the bar on the skills those require.


2. The lead's stance: six principles

Write principles like these with the team (see Design principles) and share them:

  1. Humans own the decision. AI proposes; a named designer decides and is accountable. See Human-in-the-loop.
  2. The quality bar doesn't move. AI-made work goes through the same critique, accessibility and design-system checks. See AI design review process.
  3. The design system is the grammar. AI output must use our tokens and components. Off-system output is a sketch, not a deliverable. See Design tokens.
  4. Data protection first. No customer, user or employee personal data in non-approved tools. When in doubt, ask. See AI governance and data protection.
  5. Transparency. Say when AI was used in research synthesis or generated content. Never present AI output as user evidence.
  6. Learn in public. Share experiments, failures included, in show & tell. Nobody should feel they're falling behind alone.

3. The AI-era design workflow

Step Who What happens
1. Frame Human Problem, users, outcomes
2. Specify Human + AI Design spec + context (design system, rules)
3. Generate AI, with human steering Options, prototypes
4. Review Human Critique, accessibility, design-system check
5. Refine Human + AI Iterate on the chosen direction
6. Hand off / ship Human + AI Code via the design system, annotations, QA
  • Frame stays human, and it gets more important because generation is cheap.
  • Specify is the new core skill: a good design spec (problem, users, constraints, states, rules, design-system references) makes generation useful. A vague prompt gets generic output.
  • Generate with the right tool for the fidelity you need (AI tools landscape for designers).
  • Review against the spec, principles and system (AI design review process).
  • Hand off through the design system and code, ideally with agents reading Figma through MCP (Generating design with Claude Code, Figma AI).

4. Guiding designers: the people side

Full detail: Guiding designers through AI adoption.

4.1 Expect four different reactions

Group What they need from you
Enthusiasts (already experimenting) Channel energy: make them AI champions, ask them to document and share, and set guardrails
Pragmatists (curious, busy) Concrete use cases with time saved, templates, 30-minute demos
Sceptics (quality or ethics concerns) Take concerns seriously, involve them in the review process. Sceptics make great reviewers
Anxious (fear for their role) Psychological safety, honest conversation about how roles evolve, a growth path. See Psychological safety

4.2 The AI adoption ladder (per designer, per task)

See AI adoption ladder.

  1. Explore. Try tools on low-risk tasks (rename layers, summarise notes).
  2. Assist. AI drafts, the designer edits (specs, research guides, copy).
  3. Co-create. Iterate with AI on options and prototypes within the design system.
  4. Orchestrate. Designer directs agents across steps (spec → prototype → code → review) and owns the outcome.

Use this like Situational leadership: different designers sit on different rungs for different tasks, and that's fine.

4.3 Skills that become more valuable

  • Problem framing & taste. Choosing among many options.
  • Writing. Specs, prompts, rationale. Writing is now a design tool.
  • Systems thinking. Design systems, tokens, patterns as machine-readable rules.
  • Critique. Evaluating work you didn't make yourself.
  • Code literacy. Reading generated code, understanding components and states.
  • Domain depth. Knowing your users' workflows (in logistics, finance, healthcare, wherever you work) is what AI lacks most.

→ Add these to the Career ladder and to every Individual development plan.

4.4 Protect juniors from deskilling

Juniors learn by struggling with problems. If AI does the struggling, they don't build judgement.

  • Ask juniors to explain AI output: "Why is this layout right for an operator who processes 200 orders a day?"
  • Alternate tasks done with and without AI for core skills.
  • Pair them with seniors in Design critique of AI-generated work.

5. Governance: the lead's minimum

Full detail: AI governance and data protection.

  • Approved tools list, agreed with IT security, the data protection officer (DPO) and procurement: which tools, which plans (enterprise data controls), which data classes.
  • Data rules for UX: research recordings, transcripts, customer and operational data → only approved, contracted tools. Anonymise before analysis.
  • AI literacy. Under EU AI Act Art. 4 (as amended by the 2026 Digital Omnibus), organisations deploying AI must take measures to support their staff's AI literacy. A UX team learning path is a concrete measure.
  • Check what's already switched on. Claude Design has been on by default for Claude Enterprise orgs since 15 Oct 2026; Figma AI depends on the plan. Designers may already have tools nobody approved.
  • Works council. In some countries (Germany, for example), tools that could monitor employee behaviour or performance usually need co-determination. Involve employee representatives early (as in module 09).
  • UX AI guidelines: a one-page "how we use AI in UX" (see AI usage guidelines).

6. The design system as AI infrastructure

This is where design-system strength becomes strategic.

  • AI tools produce consistent, on-brand, accessible UI only if the design system is machine-readable: tokens (Design tokens), coded components, Figma↔code mappings (Code Connect), and written usage rules.
  • A well-documented system makes every designer and every agent faster and more consistent. A messy one multiplies the mess.
  • Pitch it to stakeholders like this: "Investing in the design system is now also investing in AI productivity."

→ Links: Pattern tiers, Contribution process, Module 8: Design Systems & Standards Governance.


7. Measuring AI's impact (without surveillance)

Measure How Avoid
Cycle time Problem → validated prototype, per project (self-reported) Per-person productivity tracking
Exploration breadth # options explored before choosing Volume as a goal
Quality Design-system adoption, a11y defects, rework after handoff Assuming faster means better
Designer experience Quarterly pulse: "AI helps me do better work" (1–5) Ignoring negative signals
Learning # shared experiments / playbook entries Mandatory AI-usage quotas

Measure the team's process, not individuals. That's also what makes it acceptable to a works council or to employee representatives (see GDPR-safe tracking plan).


8. Designing AI features in internal apps

Internal tools will increasingly include AI: document extraction (invoices, forms), anomaly detection, agents that rebook orders or draft emails. The UX lead must ensure these are designed for trust, control and recovery. See Agentic design, Trust calibration.


9. A 90-day AI plan for the UX team

Phase Actions Output
Days 1–30: Listen & set guardrails Survey: who uses what for what; clarify approved tools with IT/DPO; draft AI usage guidelines AI landscape map + guidelines v0.1
Days 31–60: Pilot 2–3 pilots on real work (e.g., research synthesis, spec drafting, code prototype with design-system components); AI champions; weekly 30-min share Pilot results with time and quality notes
Days 61–90: Scale what works Playbook entries, review checklist, add AI skills to ladder, training sessions, design-system readiness backlog Team AI playbook v1 + roadmap

This mirrors a general 30-60-90 plan, and you can present it as part of yours.


10. Talking about it

You will be asked about AI by your manager, by stakeholders, by your own designers, and in any promotion or hiring conversation. Have short, honest answers ready.

"How would you bring AI into the UX team?"

"Like any tooling change: start by understanding how designers already use it and what they worry about, then set clear guardrails with IT and data protection, especially for research data. Then run a few pilots on real work and share results openly. Two things don't change: humans own design decisions, and AI output goes through the same review and design-system checks. I'd also treat the design system as AI infrastructure, because tokens, coded components and clear rules are what make AI output consistent across our apps."

"Will AI replace designers?"

"It replaces a lot of production work, not the judgement. Framing the right problem, understanding an expert user's workflow, and choosing and defending a solution become more important when options are cheap. My job is to help designers move up to that level, and to make sure juniors still build judgement."

"How do you ensure quality with AI-generated design?"

"A spec before generating, a review process after: critique against principles, design-system compliance, accessibility checks and validation with users. AI output is a draft until a designer has reviewed it and signed off."

"What about data protection?"

"Approved tools only, no personal or customer data in public tools, anonymisation before analysis, and early involvement of the DPO and, where there is one, the works council. The same principles we apply to product analytics."


11. Practice

Evidence to collect

Keep proof in your leadership log (see 12-Month Learning Plan):

  • A before/after case: spec → AI prototype → reviewed design, with what you changed and why.
  • Your guidelines draft and the feedback you got on it.
  • One example of helping a colleague adopt an AI workflow.

Read / watch

Library → section 12 · AI & the UX lead


Check yourself

Answer out loud, then open the section to compare.

  1. What are your six principles for AI in UX? → answer
  2. What are the four reactions to AI, and what does each need? → answer
  3. Why is the design system AI infrastructure? → answer

Connections

← Module 11: Workshop Techniques · UX Lead Course · 12-Month Learning Plan →

Related

Hubs: Frameworks index · Design your UX practice

Saved only on this device.

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