The Great UXR Transformation 

It’s been over three years since a new wave of generative AI redefined UX design and research, catapulting our industry from designing products around static screens to shaping dynamic, intelligent systems. Though we’re still actively wayfinding, by now we have a far clearer vision of the role UX research plays in this new landscape.

The driver of AI experiences is the model itself, which we as researchers can directly shape by applying our insights into human behavior. This moves UXR well beyond working with product teams to impacting the underlying AI technology that is powering the UX. To get there, we must embrace technical fluency and expanded methodological toolkits alongside our superpowers of human-centered reasoning and empathy.  

By translating human needs into system levers, we can define what “good” looks like and partner deeply across disciplines to make it real. This isn’t to say UX researchers are now expected to become ML engineers or applied scientists. We are, however, expected to build the foundational skills needed to shape model behavior, guide trust, and design AI-powered experiences that truly serve people.  

It also means pooling our knowledge as a UXR community to collectively evolve, envision, and meet human needs at a time when many are grappling with uncertainty and change. A living journey, what we’re sharing today reflects our current thinking and approach to making the most of this unique moment.   

Evolving to meet the moment

Today, the product isn’t just the UI; it’s a model that rapidly evolves. That changes product development rhythms and introduces new responsibilities that are a far cry from deterministic experiences, fixed screens, predictable flows, clear handoffs, and slower cycles. How is this showing up across our workflows?

• Model behavior is now a primary design material. Outputs are probabilistic, contextual, and shaped by prompts, grounding data, evaluation criteria, and finetuning.

• Product ships earlier and evolves continuously. AI experiences improve through ongoing iteration, evaluation, and data, not only UI updates.

• Success depends on understanding both human and model behavior. How people think, trust, and interpret the system directly affects how the model must behave.

UXRs are uniquely positioned to drive this impact, but if we don’t change how we operate, other disciplines won’t wait for us. When defining the places where UXR adds the most value, five areas for change emerged, each with a unique way to amplify our impact.

From purely interface-focused to model-forward research

UX researchers who can connect human insight to the levers of model behavior will shape AI experiences that deliver real value, not just sleek design or gimmicks dressed as innovation. We need to evolve past designing screens, features, or even apps and towards influencing outputs and outcomes. UXR can shape output quality through model evaluation, fine-tuning data strategy, prompt engineering, and designing UI that helps users work with the model’s strengths and weaknesses. Our work also helps guide what “good output” means for different users and contexts.

From clear boundaries to fluid collaboration

As the silos break down between traditional roles like UXR, Designer, PM, Engineer, or Applied Scientist, co-ownership of outcomes will become more common—if we open ourselves up to the possibility. The UXRs who embrace cross-functional collaboration and can stretch beyond traditional responsibilities will gain voices as true product-makers, not just late-stage advisors.

From long cycles to rapid iteration

To maximise impact, researchers need to balance disciplined prioritization with rapid execution are best positioned to deliver maximum impact during the windows when product decisions are still being shaped. We must:

• Embrace product research becoming faster, more iterative, and more tightly looped.

• Run many lightweight experiments and iterative, scoped studies, instead of fewer, larger studies.

• Leverage user feedback to shape model behavior in near-real time.

From technical awareness to technical fluency

To effectively translate research findings to shape model behavior in meaningful ways, the most successful UXRs will combine human insight with technical fluency. Essentially, we need to gain enough technical understanding of LLMs to be able to speak the same language as our cross-functional partners. This lets us:

• Frame user needs as model hypotheses and translate research findings into data or evaluation requirements.

• Contribute to decisions around prompts, tuning data, and evaluation criteria.

• Understand what’s technically feasible and what levers exist to improve the model, from design to prompt engineering to fine-tuning.

• Run faster, tighter feedback loops with model teams.

From primarily qualitative to mixed-methods

The UXRs who can flex between qualitative depth, quantitative scale, and creative experimental designs—and strategically shift between scrappy and rigorous methodologies—will answer the widest range of questions and drive the greatest impact. To get there, we:

• Merge qualitative insight, behavioral data and model metrics in our analysis.

• Become more flexible methodologists, combining qualitative skills, survey writing, experimental design, and quantitative analysis.

• Plan analysis up front to structure surveys and experiments that answer the right questions. AI can run the math, but humans must guide the choices that shape interpretation, making analytical reasoning essential!

• Know how and when to run both scrappy and exploratory or polished and rigorous research.

Turning human understanding into model-shaping direction

Human insight is our always and forever foundation, connecting computational power to humanity. Our ability to identify mental models helps teams design more intuitive prompts, flows, and affordances. And by blending creativity and rigor, we can unlock new ways of capturing model forward insights and UX data.

This lets us define the scenarios and jobs-to-be-done where model improvements matter most, grounding investment decisions in real user needs. By illuminating what people need, why it matters, and where to go next, we can bring clarity to complexity, turning uncertainty into a direction that grounds innovation in the richness of human experiences.