In the field of UX research, the rise of artificial intelligence is changing the way researchers work and think. This article explores the author’s deep reflections on the application of AI in user research, the importance of judgment and intuition, and how to translate research into practical action at Dovetail’s 2nd annual conference, Insight Out.
I always enjoy talking to UX professionals, especially those with a research spirit.
Many researchers have to deal with daily tasks with very few people who have experience or are fluent in the language of research, which can be lonely. The opportunity to gather with like-minded individuals and discuss research methodologies and trends not only broadens one’s horizons but also brings a sense of ease.
At Dovetail’s second annual conference, Insight Out, I did enjoy meeting new people (and familiar faces!). ）。 But this year’s vibe and messaging was very different from some of the research-focused conferences I’ve attended since I entered user research 13 years ago。
That’s not to say it’s bad – in fact, I found most of the conferences to be thought-provoking, and I’m glad they featured speakers from related disciplines such as product management, design, marketing, and engineering.
But many conferences hosted by user research software companies tend to focus on methodologies, case studies, and ways of working, as well as computational influence and different organizational structures.
Of course, AI has become an increasingly important part of the discussion over the past few years – and this trend continues here.
When I review my notes, I notice that the words I capitalize, bold, or circle are as follows:
- Institutions/decisions
- judgment
- Gut feeling
These are often discussed in the context of what we can’t/shouldn’t give in to AI, which is interesting in itself.
What does a product manager need to do?
In the process of a product from scratch, it is not easy to do a good job in the role of product manager, in addition to the well-known writing requirements, writing requirements, writing requirements, there are many things to do. The product manager is not what you think, but will only ask you for trouble, make a request:
View details >
But I have been reflecting and I realize that these terms are usually rarely mentioned at research-themed conferences. They are more about what research teams like me are able to achieve or demonstrate than about the research itself.
Let me elaborate.
As I said before,Research should not provide answers, but should inspire action。 I’m a little used to advocating this point of view myself. Sure, people would nod their heads in agreement, but I haven’t seen researchers actually accept it.
Dovetail just announced an interesting thing in one of its products, which is the ability to make AI-generated feature requests or product requirements for Jira or other product team tracking tools.
Telling the product team or designer what to do after completing a project often upsets traditional researchers (for some reasons), but ultimately it’s our job to help them take action – prioritize this design, put this potential backlog on hold, release this feature, and fix this design flaw.
So connecting what we’ve learned to what needs to be done next for the product/design ultimately enables the team to take the actions they need to improve the product – makeDecisions and institutionsBe able to make a difference.
This may need to be a focus for future research professionals to focus more on.
“Judgment”is another unique term that Insight Out emphasizes; Many researchers like to stick to the facts. Let the data do the talking.
But whether it’s our stakeholders or the research team itself, we have to make a lot of judgment when providing user data to our stakeholders.
For example, I have a team of only 3 researchers, but I have to support about 20 designers and more product/marketing teams. We simply can’t provide customized insights for every stakeholder.
This forces us (and our broader stakeholders) to use our best judgment when testing what and how.
For example, we have a car buying process with about 5 different product teams in charge. It doesn’t make sense to test each small part of their end-to-end process on an ad hoc basis because the online car buying experience is immature and it’s difficult for testers to get involved.
Therefore, we have established a quarterly test program that takes users through the entire car buying process (combined with field testing and prototype experience). Each designer/product/development team focuses on the part of the process they are responsible for and works on updates to incorporate them in the next test. While quarterly may not feel like frequent enough, it does ensure that everyone has access to stable user insights that inform their roadmap without us having to struggle with one-off test requests.
In addition, we make the best judgment when similar project requests arise; Instead of saying yes to every request, we sometimes help team members revisit past interviews and user testing sessions and evaluate them with new perspectives.
We will ask questions like:
- If the insights stakeholders get from users are 80% similar to “real” users, is this enough to implement?
- Can we use AI-generated summaries that cover previous rounds of research? Or do we feel the need for a data/literature review?
While our team would love to help everyone get the latest feedback, we really don’t have enough resources. So we have to make a lot of judgments, and I’ve never heard user researchers talk too much about how important this is.
This also touches on the last term that keeps popping up in Insight Out presentations:Gut feeling。
This seems to contradict the findings – we should prevent teams from making intuitive decisions, right?
But then I started thinking back to my years in the UserTesting Professional Services team, where I executed research projects on behalf of clients from all walks of life.
As a newer user researcher, I really like this because I can get a little idea of what works (and what doesn’t) in e-commerce, SaaS, automotive, hospitality, and other industries.
After a few months on the role, I began to notice elements that seemed to cause friction – inconsistent language, poorly placed CTA buttons, and unreasonable design hierarchies. I don’t always know what these elements are called, but I guess there will definitely be problems in some parts of the process.
Many times, I was right.
Because at some point, you’ve interacted with users enough to recognize what causes them to make mistakes; Your intuition kicks in, and you become more aware of their needs and how they behave.
Whether researchers like it or not, we have it, and so do our stakeholders – those who take the time to listen to users, anyway.
Of course, the most prominent topics throughout the conference were:artificial intelligence。
Overall, UX research disciplines seem to realize that they need to embrace its use and “learn to dance with the devil,” as one panelist put it.
We had high hopes for AI when it came to collecting and managing what another panelist called “the most valuable company intellectual property – customer knowledge.”
Dovetail even shared a *concept* futuristic design in which your customer data will be transformed into a human-looking avatar that allows you to talk to your own customer knowledge as if it were a person.
Based on my conversations throughout the day, audiences have not reacted very well to this, as has the idea of using AI to conduct surveys or interviews on your behalf (I suspect because for most researchers, connecting with users is one of the best parts of the job).
At the moment, the field seems to be rightIncorporating AI into the analysis and reporting phase of our workBe more open-minded. It does a good job of spotting topics, but doesn’t help much in more generative work – the AI can’t tell you where to dig, or exactly what questions to ask, especially when it comes to entirely new experiences.
Personally, I find that there are some other gaps in AI-backed research. There are many situations that researchers need to deal with, and I’m not sure AI can solve them at the moment, such as:
- captureObserve the data(What people click, where they hover, which page they are on, facial expressions, body language)
- captureFalse positive(When testers confidently claim they know exactly what something means or how something behaves, but we know they are completely wrong)
- I don’t understandTerminology or context within the team(Page/feature name that is not intended for the public, page or feature that the team is most concerned about or trending at the moment, etc.)
- Identify or address “unsuitable” situations – during the screening processFalsely state your own participants(We’ve all seen the kind of people who get into the test by talking nonsense)
- copeIndifferent, sarcastic, or sarcastic(“Whatever” or “your mother” comments, “that’s great”, etc., when they mean the opposite)
Maybe I’m not aware of the latest and greatest AI augmentation technology (it feels like everyday chore), but it seems like the technology needs to evolve a lot before users research actually automating it to replace people’s jobs.
So, how do I feel after attending Insight Out 2025?
Well, maybe because of the red-eye flight that spans 3 time zones, but even after a week, I still feel a littletired。
As people outside of tech startups/fast-growing companies, we feel like we’re adapting too quickly.
I would love to see and hear more practical guides, in-depth use cases, or case studies on how AI can effectively enhance our work – while there’s a lot of talk about how AI can “automate” research, writing high-quality prompts and plans for your AI tool/assistant feels important to do it well. I guess I’m not the only one who is at a loss as to how to dance with this devil.
More broadly, the message I get is that it may be time for a user research professionalGo beyond tools and methods.In the era of rapid change, empower and demonstrateDecision-making, judgmentandIntuition can make us better researchers and team members.
I look forward to participating again in 2026!
