UXLx 2025 — Wrap Up — Talks Day
23 May marked the final day of UXLx 2025 — and that means Talks Day! 🤩 10 industry leaders took the stage to share insights with an audience UX professionals coming from 33 countries, all eager to learn and recharge their creative energy.
The talks spanned some of the hottest topics in UX today, including AI, design systems, accessibility, research, design ops, and content.
This wrap-up is your chance to revisit the key takeaways if you were there — or, if you couldn’t make it, to catch a glimpse of the conversations shaping the future of UX.
🤝 UXLx thrives on collaboration, and this article is no different — jump in, highlight your favourite moments, and share your key takeaways!
💡 Read through the article or click a talk title to jump directly to the one you’re interested in.
23 MAY
🕘 MORNING
- Themeable Design Systems with Design Tokens, Brad Frost
- Patterns for Faster, Cheaper, Better UX Text, Torrey Podmajersky
- Preparing Teams to Design Impactful AI Experiences, Mike Oren
- Design for the rest of us — Must-haves of accessible design, Devon Persing
- DesignOps Initiatives in Practice, Peter Boersma
🕑 AFTERNOON
- AI is Transforming UX Research, John Whalen
- Design for Impact — Rally Your Team Around a Process that Drives Growth, Erin Weigel
- Designing for AI Agency — How AI Agents are Redefining Customer Experience Design, Matias Vaara
- Leveraging Cultural Knowledge for Global UX, Chui Chui Tan
- UX vs AI — Fight!, Scott Jenson
Themeable Design Systems with Design Tokens by Brad Frost
The problem: “Multi-all-the-things” organisations
Modern organisations often manage multiple brands, products, platforms, and technologies, each with distinct audiences, goals, and interfaces. This creates complexity in maintaining visual and functional consistency.
Examples:
- Multi-brand: Marriott (many hotels with distinct styles)
- Multi-product: Caterpillar (e.g. cat.com, parts.cat.com)
- Multi-platform: Web, mobile apps, kiosks, POS systems
- Multi-framework: React, Angular, Drupal, etc.
The challenge with change (e.g. rebrands)
- Changing a core design element like a colour (e.g. “Starbucks green” to “Starbucks purple”) is painful and expensive.
- Every team must update their implementation separately: duplicated, disconnected, slow, error-prone, and very costly.
The solution: design tokens
Design tokens are centralised, abstract variables that define design properties (colours, typography, spacing, etc.) and are used across platforms. They:
- Act as a shared language between design tools (e.g. Figma) and code.
- Enable themeability and fast global changes.
- Bring efficiency, consistency, and scalability.
When the design token “brand background” points to “Starbucks green,” every product using that token automatically reflects a brand change when the value is updated.
How it works: separation of concerns
- Inspired by web development’s separation of HTML, CSS, and JS.
- UI components (e.g. buttons) define structure and behaviour.
- Design tokens define style, which can be swapped without changing structure.
Token tiers: a scalable token strategy
Brad outlines a three-tier design token architecture:
- Tier 1 — Definition: Fundamental, reusable values (e.g., #00704A for green). Not shared publicly.
- Tier 2 — Semantic Usage: Assign purpose (e.g., colour-background-brand) and connect to UI context.
- Tier 3 — Component: Used sparingly for special cases (e.g., button-primary-background).
This allows design systems to support:
- Dark/light modes
- White-labeled products
- Localised UI
- Rebrands and refreshes
Components + Tokens = Flexibility
- UI components (like buttons) can remain structurally consistent while changing style via tokens.
- Multiple themes can be applied (e.g. marketing vs. enterprise themes).
Why it matters
Design tokens empower organisations to:
- Scale design systems across teams and platforms.
- Achieve fast, consistent rebrands.
- Enable efficient collaboration between design and development.
- Reduce costs and risk.
Final thought
Design tokens represent a new foundation for design systems — enabling themeability, consistency, and agility in “multi-all-the-things” organisations.
Patterns for Faster, Cheaper, Better UX Text by Torrey Podmajersky
Good UX content isn’t just opinion
- Early web design let anyone choose fonts, colours, and words — often arbitrarily.
- Today, content decisions should be systematic, measurable, and aligned with product goals.
UX content is content for doing
- It supports the user journey: attracting → converting → onboarding → engaging → supporting → transforming.
- It’s distinct from marketing or branding content — focused on helping users act, decide, and succeed.
Design teview bottlenecks
- Content often slows down design reviews, even when visual elements are approved.
- Using consistent patterns (like design tokens or voice guidelines) helps teams avoid repetitive debates.
UI text patterns
Specific patterns for different UI elements:
- Context Labels: Use recognisable keywords and names people expect.
- Action Items: Follow “verb the noun” pattern (e.g. “Create an account”).
- Menu Items: Can mix nouns and verbs but design as cohesive sets.
- Descriptions: Keep to 3 lines maximum, avoid asterisks, place keywords strategically.
- Empty States: Explain what’s missing and how to fill it.
- Transitions: Show process status (e.g. “saving” → “saved”).
- Error Messages: Remove unnecessary words like “please”. Describe the problem and provide clear next steps.
- Blocking Errors: Explain what happened and what users can do (“You can’t play until you reconnect.”).
The patterns serve as starting points for iteration, not rigid rules. Embedding them in design components helps teams start from consistent foundations while maintaining flexibility for specific contexts.
Preparing Teams to Design Impactful AI Experiences by Mike Oren
AI is the next big disruption for UX
- Just like the rise of the internet and mobile, AI — especially generative AI (GenAI) — is transforming how we design.
- Designers must go beyond simply adding prompt boxes and embrace deeper integration with back-end systems and data.
Designers must collaborate across disciplines
- UX now requires working closely with data scientists, machine learning engineers, and understanding data pipelines.
- A key challenge is aligning what AI can do with what users expect it to do.
Understand and work with AI limitations
- “Hallucinations” (AI outputs that seem wrong) are part of how GenAI works, not flaws.
- Designers can reduce or leverage hallucinations by shaping input, setting context, and collaborating with engineers.
- Clear user expectations are essential — use terms like “draft” instead of “create” to communicate that human input is still needed.
Data literacy is crucial
- Designers don’t need to be data experts, but should understand basics like linear regression, mean vs. median, and data distributions.
- This helps them contribute meaningfully in decisions that affect AI behaviour and user experience.
Design hiring must evolve
- Design leaders should look for candidates (or upskill existing team members) with data fluency, AI awareness, and collaborative mindset.
- Service designers, behavioural designers, and motion designers are particularly valuable for AI-rich experiences.
Motion and behavioural design are underutilized
- Motion design can build trust and convey AI’s “thinking” visually.
- Behavioural designers help users adopt AI by making experiences feel less intimidating and more magical.
Service design mindset is essential
- Designers must think holistically — across touchpoints, platforms, and systems.
- AI should be integrated meaningfully, not just added on superficially.
Embrace discomfort and experimentation
- Designing with AI is messy, evolving, and uncertain — just like web and mobile were in their early days.
- Success will come from cross-functional collaboration, experimentation, and specialisation.
Move beyond commoditised design
- Much of product design has become repetitive and solved; AI opens new frontiers.
- This is a chance to rekindle innovation and rethink user experiences from the ground up.
Design for the rest of us — Must-haves of accessible design by Devon Persing
Accessibility is a design issue, not just a technical one
- Accessibility is often treated as a technical fix at the end of a project.
- Instead, it should be integrated from the beginning — during ideation, design, and prototyping.
Understanding disability
- Disability is context-dependent and often shaped by society’s failure to accommodate people.
- The social model of disability sees barriers as created by environments, not individuals.
- Accessibility must consider: Perception, Interaction, and Cognition.
- Disabilities can be permanent (like osteoarthritis), temporary (like having the flu), or situational (like breaking glasses).
- Co-occurring conditions are common (comorbidity). These lead to a variety of fluctuating symptoms that affect digital interactions.
Accessibility affects everyone
- It intersects with every demographic: age, race, gender, economic class, geography, etc.
- Important to design inclusively with this diversity in mind.
- 78% of WCAG 2.1 criteria have an interaction content or visual design component — not just development.
- Accessibility is about context, not just checkboxes.
Design fails and real-world examples
Motion and Animation:
- Issue: Excessive motion (e.g. parallax scrolling, autoplay videos) can trigger migraines, vestibular disorders, or overwhelm neurodivergent users.
- Fixes: Respect system preferences for reduced motion. Avoid autoplay. Provide clear, accessible pause controls.
Form Workflow:
- Issue: Poor information hierarchy and unclear actions (e.g. hidden “Send Code” buttons).
- Fixes: Logical reading and interaction order. Clear instructions. Minimise user memory load.
CAPTCHAs and Security:
- Issue: CAPTCHAs are often not accessible (e.g. image grids, timed challenges).
- Fixes: Avoid using CAPTCHAs that require visual or rapid motor input. Use backend bot-detection methods instead of user interaction challenges.
Practical advice
- Include accessibility in early stages of the design process.
- Test with users with disabilities.
- Use tools (like Figma plugins) to annotate accessibility requirements.
- Be proactive — assume disabled users already interact with your products.
- Hire disabled professionals for design and research roles, not just advocacy.
“And also just please be kind to your users and to each other.”
DesignOps Initiatives in Practice by Peter Boersma
“We continuously and structurally improve the circumstances for designers and the design org, to increase the chances of good design happening.”
DesignOps overview
DesignOps specialists take on responsibilities traditionally held by design managers:
- Defining and agreeing on design processes
- Work tracking
- Designer training
- Design culture development
- Tool selection and implementation
- Meeting/event coordination
What is DesignOps?
DesignOps (Design Operations) focuses on:
- Improving the processes, tools, and culture supporting design teams.
- Allowing design managers to focus more on creative direction and talent development, while DesignOps specialists handle operational tasks.
Why DesignOps?
- Designers and design leaders benefit from smoother workflows, clearer processes, and better tooling.
- Continuous improvement mindset: small, ongoing changes (even 1% daily) lead to big gains over time (38x improvement annually).
Examples of DesignOps initiatives
1. Design Review Team (at ServiceNow)
- Large-scale design org (275 people) supported by a 25-person DesignOps team.
- Established review checkpoints in the product development lifecycle.
- Included accessibility, design system, and conceptual design checks.
- Enforced by a “design police” metaphor — projects couldn’t proceed without approval.
2. Documented Design Process (at Miro)
- Mapped the product development lifecycle: Discover → Define → Design/Research → Build → Ship.
- Created detailed documentation of goals, deliverables, and collaborators per stage.
- Included a tailored version for growth teams with different workflows.
3. Horizontal vs. Vertical Work (at IKEA)
- Horizontal = org-wide initiatives (e.g. career ladders).
- Vertical = team-specific support (e.g. helping one design team with tailored processes).
- Created a “DesignOps partner” role for vertical support, on a request basis.
Implementing DesignOps initiatives: 5-step framework
1. Inventory: Collect a broad list of potential initiatives (e.g. from team pain points).
2. Prioritise: Rank by urgency, impact, alignment with company OKRs. Place initiatives in a Now/Next/Later plan organised by themes (e.g. tools, people, process).
3. Distribute: Distribute the work because you can never do it alone.
4. Coordinate: Write briefs and assign responsibility across the team.
5. Measure: Track current state vs. desired outcomes to evaluate impact.
Key takeaways
- DesignOps is about improving the system around design, not design itself.
- Initiatives should be intentional, documented, prioritised, and measured.
- The work is collaborative and evolves with organisational needs.
- Gartner prediction: Teams with DesignOps support increase revenue at 2x the rate of competitors.
AI is Transforming UX Research by John Whalen
AI is revolutionising UX research by dramatically increasing speed, scale, and efficiency, shifting researchers’ roles from manual work to higher-level analysis and strategic impact.
Key research transformation insights
- Initial skepticism about AI was challenged when AI tools produced 80% of the insights human researchers did.
- Even seasoned researchers were surprised at how capable AI interviewers and assistants have become.
- John’s team at his company Brilliant Experience ran comparative studies: traditional human-led interviews; Human + AI-assisted research; Fully AI-moderated interviews; Fully synthetic studies (AI-generated participants and data).
- Average study duration dropped from 7 weeks to 7 days — a reduction of 85% — with similar quality results.
- AI now handles logistics, scheduling, and even moderation, allowing researchers to focus on analysis, storytelling, and stakeholder engagement.
AI tools demonstrated
Synthetic users are AI-generated personas that can:
- Simulate target users.
- Be used to test interview questions.
- Provide input in Jobs-to-Be-Done format.
- Represent different demographics or behaviours for global scalability.
AI-moderated interviews
- Tools like Listen Labs allow AI to conduct dynamic interviews, adapt based on responses, and provide real-time analysis.
- Can conduct interviews in multiple languages.
AI-powered analysis
- Tools like CoLoop let researchers query transcripts, generate summaries, and verify findings directly from data.
- You can chat with the data, pull quotes, and access relevant clips instantly — enabling “trust but verify.”
Global, scalable research
- AI moderators work across time zones and languages — e.g. running 300 interviews over a weekend.
- Enables more inclusive, multilingual, and cost-efficient studies.
Changing roles for researchers
- Researchers move from “doers” (moderating, transcribing) to “orchestrators” (guiding tools, interpreting insights).
- AI doesn’t replace researchers — it augments their abilities and expands their strategic value.
Call to action
- Learn prompt engineering — it’s vital to get quality output from AI.
- Be ready: these tools are evolving rapidly. Staying ahead requires experimentation and upskilling.
- Take courses, try tools, and be curious about the potential of AI in research.
Predictions for the future of UX research
- Cost of research will keep decreasing.
- Limitations around scale and global reach are dissolving.
- Researchers must adapt to stay relevant, leveraging AI for greater impact.
Design for Impact — Rally Your Team Around a Process that Drives Growth by Erin Weigel
Core concept: conversion design
Conversion derives from the Latin “convertere” which means to transform or change. Conversion design is the practice of crafting intentional change that leads to measurable improvement, not just difference.
It’s grounded in design (intentional outcomes), science (experimentation & evidence), and business (value creation).
Systems thinking vs. Linear thinking
- Most teams use linear or loop-based processes (e.g. discover-design-develop-deliver).
- Conversion design emphasises systems thinking, acknowledging messy, nonlinear progress and complex interdependencies.
The conversion design process (7 steps)
- Understand: Research problems through a business goal lens.
- Hypothesise: Formulate educated guesses (fancy guessing).
- Prioritise: Prioritise the guesses.
- Create: Design and build potential solutions.
- Test: Run experiments (ideally A/B tests) to validate hypotheses.
- Analyse: Interpret the results and data.
- Decide: Use evidence to make well-informed decisions.
The output: Generate collective knowledge and business value.
Business impact: value & growth
- Conversion design aligns experimentation with business goals and customer problems.
- Teams work to increase customer value, which powers the value cycle: business → employee → task → customer value.
- Driving growth requires cultural change: not just one team experimenting, but everyone doing it.
Hierarchy of evidence (from lowest to highest reliability)
- Expert opinion
- Observational study
- Randomised Controlled Experiment
- Systematic review (best: combines all evidence with context and human judgment)
Case Study — A/B Testing
- Context: Testing accessibility improvements for mobile app
- Change: Enhanced contrast ratios in UI to meet WCAG guidelines
- Initial Mobile Test Results: Significant increase in sales
- Web Implementation:
- Team suggested skipping A/B test
- Testing revealed implementation bugs causing profit loss
- Demonstrated importance of testing every implementation
- Prevented widespread deployment of problematic code
Lesson: Always test — the same concept can have different results due to execution differences.
Takeaways
- Conversion design combines design, science, and business.
- Use your whole brain — linear + systems thinking.
- Run well-designed experiments.
- Create a culture of evidence-based iteration.
- Ultimately: aim to make things better, not just different.
Designing for AI Agency — How AI Agents are Redefining Customer Experience Design by Matias Vaara
What is agentic AI?
- Agentic AI goes beyond generative AI (which answers prompts) to take action on behalf of users.
- Think of it as moving from suggestion (e.g. restaurant lists) to execution (e.g. booking a table).
- Enabled by improvements in:
- Reasoning (Powerful LLMs that can plan and orchestrate actions)
- Context (Real-time access to user, org, and web data)
- Tools (Secure API & UI control to do real work)
The problem it solves:
- Modern users juggle too many tools and workflows — similar to manual switchboard operators in the 1890s.
- Agentic AI promises automation of digital grunt work, freeing people for strategic and creative tasks.
Three early principles for designing agentic AI
1. Useful
- Focus on real, tedious problems, not tech gimmicks.
- Prioritise value-first, harm-never.
- Examples: Zendesk AI resolving 60% of tickets; MoveWorks handling HR/IT requests; Zip AI agents helps with tedious work in procurement.
- Avoid scaling harm — build in guardrails.
2. Collaborative
- Stop thinking in deterministic UX flows — start thinking about AI teammates.
- Agents should:
- Act like executive assistants — anticipate, clarify, and act.
- Be proactive yet helpful (e.g. AI sales agents, product recommenders).
- Be transparent — users should know what agents are doing and why.
- Handle multimodality (text, voice, etc.) and seamless handovers between AI and humans.
3. Controlled
- Users must remain in control:
- Critical actions should require human approval (e.g. GitHub Copilot code changes).
- Provide intuitive control panels with appropriate granularity.
- Design for mandatory engagement to avoid “over-delegation.”
- Build trust through transparency, privacy, and fairness audits.
Future guidance
- AI agents are coming fast — designers must actively shape how they’re built and used.
- Start with your own workflows: identify tedious tasks and experiment with tools like Lindy AI.
- Stay involved to bring humanity, ethics, and usability to AI agent design.
Leveraging Cultural Knowledge for Global UX by Chui Chui Tan
Understanding and integrating deep cultural knowledge — beyond language and interface translation — is essential to creating globally successful user experiences and business strategies.
Cultural context shapes user behaviour
- During a study, a participant from the Philippines asked Chui Chui’s age to address her respectfully — highlighting how cultural norms affect interactions.
- Misinterpreting such behaviours without context can lead to poor UX insights and decisions.
Failures due to poor cultural fit
- Several global brands failed in certain markets because they didn’t align with local consumer values, behaviours, or expectations. Examples:
- Sephora: Misalignment with South Korea’s skincare-focused and subtle makeup preferences.
- Walmart: Couldn’t adapt to German retail culture.
The evolution of localisation
- Stage 1 — Pre-2000s: “One-size-fits-all” mindset.
- Stage 2 — Early 2000s: Surface localisation — language, currency, formatting.
- Stage 3 — Mid-2000s: Functional localisation — local payment methods, social platforms.
- Stage 4 — Now: Deep cultural adaptation — integrating cultural psychology and local ecosystems.
“Culturalisation” vs. “Localisation”
- Localisation is often seen as superficial.
- Culturalisation involves continuous, layered adaptation that accounts for evolving cultural, economic, historical, and psychological factors.
Culturalisation case studies:
- IKEA in India: Visited 1,000+ homes across cities to understand living conditions. Adapted products, store layout, café menu (samosas, vegetarian meatballs), and pricing strategies to local lifestyles and physical environments.
- Bumble in Japan: Analysed local dating norms, app features, pricing expectations, and gender roles to adapt “women make the first move” to a culturally sensitive and successful experience.
- Spotify in the Philippines: Introduced short-term subscription plans inspired by sachet-buying culture rooted in economic behaviour and history.
Cultural research Is complex
- It’s not just about hiring a local translator or agency.
- Cultural research requires deep immersion, understanding of societal structures, historical influences, and sometimes adjusting incentives (e.g. luxury hotel guests in Qatar valued status over monetary compensation).
Toolkit and frameworks
- Chui Chui developed a Cultural Activation Toolkit with 27 cultural aspects to help teams identify what matters in a given market.
- This helps avoid relying on assumptions or stereotypes when entering or growing in a new market.
To succeed globally, businesses must move beyond superficial localisation and commit to continuous, in-depth cultural understanding — across design, marketing, pricing, product, and business strategy. It’s not just a design team’s job — it takes the whole organization.
UX vs AI — Fight! by Scott Jenson
From excitement to frustration: the AI hype cycle
- Scott describes an emotional journey: from early excitement with tools like ChatGPT to deep skepticism and confusion.
- The tech industry repeatedly goes through hype cycles with new technologies. Gartner’s hype cycle shows how technologies move from trigger to peak of inflated expectations, through a trough of disillusionment, before reaching productive use. Historical examples include MOOCs (2011) and smartphones (1999–2009). Current AI investment exceeds $300 billion with less than $2 billion annual revenue, suggesting an approaching market correction.
- Critiques naive adoption, reminding us that culture and user behaviour often lag behind technological advances.
LLMs and the illusion of intelligence
- Argues LLMs like ChatGPT don’t truly “understand” or “summarise” — they shorten, based on patterns in their training data.
- Explains anthropomorphism: we instinctively assign human traits to AI (e.g. “it’s hallucinating”), which skews how we interact with it.
The role of UX in making sense of AI
- Warns against simply adding chatbots or AI as an afterthought — “slamming a DOS prompt into your app”.
- Shares case studies where small, invisible uses of AI (like query rewriting) had outsized impact.
- Emphasises the need for structure, context, and user understanding in any AI integration.
Anchors for moving forward
Scott offers four practical “anchors”:
- Anchor 1: Look at your users first — Start with users and business needs — not tech hype.
- Anchor 2: Learn from others — Especially failures and academic research.
- Anchor 3: Add structure to the conversation — avoid raw, vague prompt-driven UIs.
- Anchor 4: Transmogrify your data: Understand your data — not all content is equally AI-friendly.
Outlook: prepare for the trough
- Believes we’re nearing a “trough of disillusionment” for AI, similar to past cycles.
- Encourages UX professionals to play an active role in shaping the future — challenging hype, grounding in human needs, and guiding design ethically.
“We’re not anti tech. We’re just trying to get ready for this thing to figure out what can happen, how we can make these products better.”
After party — IDEA Spaces rooftop 🎉🍻😎
After soaking up all the knowledge from UXLx 2025’s workshops and talks, it was time to celebrate! 🎉 This year, the party took place at the stunning IDEA Spaces Rooftop — courtesy of our amazing Gold Sponsor.
With food, drinks, live DJ, and panoramic views over Lisbon, the vibe was unbeatable. ✨ The perfect setting to wrap up four unforgettable days of learning and connection.
See you next year? 😉
👀 Stay tuned!
We’ll start releasing the videos from the Talks Day on our UXLx Videos page, gradually, when we launch the next UXLx event. In the meantime you can check hours of content from the previous editions.
📩 Sign-up to our mailing list to get the latest UXLx news and updates.
📸 All photo credits go out to UXLx’s official photographer — José Goulão.