AI as UX Research Collaborator 🫶 not Competitor
Some semi-spicy🌶️ reflections on the future of UXR roles
Earlier this month, I took John Whalen's course AI Skills for Research: Create an AI-Powered Insights System on Maven. John and his team at Brilliant Insights have been kicking the tires on AI Research tools since the early days, openly sharing their findings with studies like AI vs. Human: A User Research Showdown.
The two-week course was a valuable sampler on how you can use AI end-to-end, across your research projects. I especially enjoyed getting hands-on and trialing tools with real participant data.
The course helped me further reframe my perspective on AI—from seeing it as a potential competitor to embracing it as a powerful collaborator for UX Researchers.
As part of this, I’ve had a few semi-spicy 🌶️ reflections on the future of UXR.
Let's dive in.
Becky’s Spice Scale
🌶️ Mild Salsa – Thought-provoking but unlikely to ruffle feathers.
🌶️🌶️ Jalapeño Surprise – A bit of a kick, raising eyebrows but still digestible.
🌶️🌶️🌶️ Ghost Pepper Panic – Spicy enough to make UXRs shift uncomfortably in their seats.
🌶️🌶️🌶️ 🔥 Carolina Reaper Chaos – So hot it induces existential dread.
🌶️ AI is the next phase of research democratization
Several years ago, UXRs hotly debated research democratization. Can 'non-researchers' really do proper research? Should we 'allow' them to do this? What does this mean for our jobs?
In some organizations where democratization wasn't officially sanctioned, PMs and Designers continued to conduct their own research, just in secret. Obviously, this wasn’t ideal. Eventually, most UXRs got on board with democratization, and now empowering People Who Do Research (PWDR) across our orgs is just part of the gig.
AI is the next phase of democratization. It's a foundational technology, spanning across the research toolbelt that can't be ignored. However you feel about AI, undoubtedly PMs, Designers, and others in your org are already using AI for research purposes — prompting ChatGPT with all kinds of things — whether you know it or not.
Even if UXR teams are still figuring AI out themselves (and really, aren’t we all? 😛), we have an opportunity to teach PWDRs how to effectively use AI for research. We must create best practices and share them, openly empowering others to use AI - not blocking them.
One easy place to begin is teaching effective LLM prompting to generate research plans and interview guides, ideally aligned with company templates.
UXRs also have a responsibility to educate PWDR about the ethical implications of AI. We must ensure that internally, AI tools are used responsibly, including how data is stored, and that research participants understand how their data will be processed (informed consent).
🌶️🌶️ AI analyzes like a junior-level researcher. And for some studies that’s enough
Though I was initially skeptical of AI research analysis tools (like Marvin and Coloop), I now realize how impressive they are. AI can dramatically reduce analysis time for studies led by UX Researchers or PWDRs. It can summarize everything participants said, find themes, and prioritize findings. Some tools can even write a report, pull quotes, and create video clip reels. Incredible… and a bit scary!
In my opinion, AI analyzes like a strong junior researcher. It’s great at (very quickly) summarizing and surfacing the “what” of a study. But, it won’t fully deliver the “so what” or “now what” (more on that below). However, just with the out-of-the-box features, these tools can get teams ~70% of the way there.
And in some cases, 70% might be good enough. That might be enough for teams to take the insights, make the decisions they need to, and move forward.
Hard to stomach? Possibly. But it shouldn’t be.
Consider this: How many projects at your org never get user insights because the UX Research team is overloaded? How often do PMs and designers want to do research but feel they don’t have time? Or worse—have you ever had teams carve out the time to conduct interviews, but then never actually analyze them? 😭
AI analysis can empower product teams to move faster with real user insights. And while it may not be as rigorous as a researcher-led analysis, I feel this is far better than no research at all.
🌶️🌶️ UX Researchers can drive value by upping their strategic skills and hybrid analyses
AI can quickly surface the “what” in research, but it struggles to define the “so what” and “now what”. Without deep product knowledge, awareness of competitors, or an understanding of company strategy, AI can’t extract the implications of insights the way an experienced UX researcher can (at least, not yet).
This shift means researchers need to sharpen their strategic skills. The insights in our reports must go beyond being merely interesting; they need to be actionable, prioritized, and directly tied to business goals.
And, with AI reducing time spent on analyzing our primary data, UX researchers can strengthen their studies with hybrid analyses—layering in multiple mixed data sources for richer insights.
This could include reviewing and interrogating (with AI):
Past UXR studies
Quantitative surveys
Product analytics
Insights from cross-functional teams (Marketing, Customer Support, Sales, etc.)
External research looking at the competitors and market
and more!
By intentionally gathering, interrogating, and synthesizing diverse data sources into a compelling narrative, UXRs can elevate their strategic impact within their organizations.
UXRs can also proactively apply this hybrid analysis approach to studies they haven’t been explicitly asked for, thinking more holistically about the insights their company needs to succeed.
This shift allows them to move from reactive to proactive, addressing broader organizational needs. By deeply thinking about what data to bring together, then thoughtfully interrogating it, and weaving it into a powerful narrative, UXRs can continue to create deep strategic value for their orgs.
🌶️🌶️🌶️ AI moderated studies means rethinking the UXR role as user-business connector
Just as tools like UserTesting.com made unmoderated research more affordable, faster, and scalable globally, AI is doing the same for moderated studies, with tools like Wondering and ListenLabs.
Of course, AI moderation isn’t as good as a trained researcher—yet. But there are also real upsides. Think about it: AI can conduct research at any hour, in any language, with multiple participants at once. It never sleeps or takes breaks. And, for some PWDR, fear of interviewing was a real barrier to gathering insights.
You might protest: But moderating is my favorite part of research! Getting teams to observe interviews is how I connect with users and help teams build empathy. That’s valid. If researchers (and product teams) only read AI-created summaries, from AI-moderated studies, never connecting with live humans, there’s a risk of becoming disconnected.
But what if we saw this shift as an opportunity rather than a loss? With AI moderating some studies, what if we think more strategically about how, when, and where we spend our time with users?
Product teams often say that real-world interactions—meeting users in their homes or workplaces—are the most powerful empathy builders. For example, Canva’s CTO, Cameron Adams, still vividly recalls meeting one of the product’s first users over 10 years ago: a small business owner designing swimwear from her garage.
Instead of seeing AI moderation as replacing us, what if we use some of that time and energy formerly spent moderating (and scheduling, reviewing transcripts, etc.) and put it towards creating more impactful, real-world "wow moments”? The kind that leaves lasting impressions on product teams, and deepen user-business connections.
🌶️🌶️🌶️🔥 Synthetic users are actually… useful?
Before taking this course, I outright dismissed synthetic users, basically believing they were the end-all-be-all of our craft. But after hands-on testing—and even creating my own—I’ve nearly done a complete 180. Let’s say… a 150.
While synthetic users are not a replacement for real participants, I now see them as a powerful complement to traditional research methods—when used appropriately.
For example, you could:
Use synthetic users during early-stage research workshops with stakeholders, to better refine research goals before kicking off studies
Run interview questions through a synthetic user to spot gaps, iterate on phrasing, and maximize the depth of insights we get from real participants—especially when recruiting niche or hard-to-find audiences.
Junior researchers or PWDR can sharpen their interview techniques by practicing with synthetic users before speaking with real participants.
When timelines are tight or participants are scarce, synthetic users can serve as an initial testbed for concepts. While they shouldn’t replace real users, they can help fill early-stage gaps and be cross-validated later with actual participants.
Offer research-based synthetic users as a more dynamic, interactive research deliverable than personas
As John puts it in the course: “Synthetic users are only going to get better; they’re the worst they’ll ever be right now.” Rather than resisting synthetic users entirely, how might we get ahead of the curve and use them to our advantage?
In Conclusion
AI can already add value across much of the UX Research workflow. While seemingly scary, we need to embrace the fact that our roles are evolving. We need to view AI not as a competitor - but as a collaborator. 🫶
Think of AI as the research intern you always wished you had. By offloading some parts of our role, UXRs can scale ourselves and our teams, and have a greater impact. We can help PWDR get more reliable insights. We can dedicate more energy and focus on higher-priority research projects ourselves. We can curate and craft deeper strategic insights, and strengthen our role as the bridge between users and the business.
I’m excited about the possibilities of this future. I’d love to hear your thoughts, in the comments!
A huge thanks again to John Whalen, his team, and his course for helping plant, progress, or cement these ideas. Sign up for the next cohort here.