At a Glance: Gabija Alčauskienė
Current Role: Senior UX Researcher, NordVPN | Nord Security
Location: Vilnius, Lithuania
Professional Background: UX Research, Market Research, Data Analysis, and Psychology
Previous Experience: Project Manager and Senior Market Research Analyst at Civitta
Education: Master’s Degree in Clinical Psychology, Vilnius University
Additional Training: Data Analytics, Turing College
Areas of Expertise: Qualitative Research, Quantitative Research, User Experience, and Customer Insights
AI Applications: Research Automation, Survey Analysis, AI-Assisted Interviews, and Research Reporting
Biggest AI Benefit: Reducing Usability Research Reporting from One or Two Weeks to One or Two Days
Research Philosophy: Human Judgment and Real User Feedback Remain Essential
Responsible AI Priority: Data Privacy, Research Accuracy, and Human Oversight
"Before, we used to need a week or two weeks to come up with the report. Right now we can do it in one to two days."
AI has cut NordVPN’s usability-study reporting from one or two weeks to one or two days, but its researchers still refuse to hand over the parts that need judgment. Kamilė Jokubaitė, co-founder of Attention Insight, spoke with Gabija, a senior UX researcher at NordVPN, about where AI helps, where it doesn’t, and where her team draws the line.
Gabija works on the NordVPN product within the wider Nord family. Put plainly, her job is talking to users and potential users: learning their needs and pain points, checking whether the products work for them, and shaping how those products should develop.
Gabija Alčauskienė brings together a background in psychology, market research, and data analytics to understand user behavior and support informed product decisions. Before joining NordVPN, part of Nord Security, she worked at Civitta, progressing through market research roles to project management, and gained experience in psychological research at Vilnius University. With a Master’s degree in Clinical Psychology and additional training in data analytics, she combines an understanding of human behavior with analytical expertise in her approach to UX research.
What AI changed, and what it didn’t
The biggest gain is time. Gabija joined a year ago, a period when AI tools “grew exponentially.” Because the team works mostly with qualitative data, the manual work of transcribing and combing through transcripts used to eat hours.
“We have more choice on what to do with our time,” she says. “We can either do more research, or we can refine.” People say AI degrades output quality, but for her team it frees time to add more human touch: sharpening insights, improving how they are presented, and building relationships with stakeholders. Those relationships matter because the team runs both requested studies and proactive research of its own.
The largest shift is in processing and analysis, especially for tactical usability studies with clear questions and clear successes or failures. The team is automating these with skills and agents. “Before, we used to need a week or two weeks to come up with the report. Right now we can do it in one to two days.”
Three things have not changed:
- Business context. AI can’t yet be fully onboarded to it, and it matters most when planning research, raising questions and forming hypotheses.
- Stakeholder relationships. Insights only have impact when the people receiving them trust the team.
- Finding hidden gems. AI leans on common patterns that many users mention. Researchers can spot the niche, surprising thing one or two users said and, knowing the business, decide it is worth exploring further.
Making sense of feedback at scale
Gabija’s team mostly works with survey responses. Closed questions go straight into dashboards; the open-ended answers are where AI earns its place. The team uses it twice: first to create the categories, then to sort every answer into one, turning qualitative responses into numbers.
They used to run this through a proprietary Python script that worked in batches. Early chat tools had small context windows, and after a few dozen responses they would invent new categories or hallucinate. Now, because the task comes up only occasionally, regular prompting in Nord’s internal tool, Nexus, is enough. “You just drop in your answers, you create categories, you get the categorization within minutes.”
The next step is automating the feedback forms on Nord’s products and websites, such as pre- and post-purchase surveys. The goal is a monthly report that categorizes open answers and flags the key changes in user sentiment.
Where to draw the line
Synthetic users are where Nord draws it, for now. The demos the team has seen produce generalized, common-sense answers. As Gabija sees it, synthetic users mainly solve the cost and time of recruiting real people, and that isn’t Nord’s problem: “Our users are really eager to share, and they appreciate being asked.”
Where AI does add value is in talking to data the team already has. Numerous studies have left hundreds of files, and researchers query them to predict how users might react to a new solution, design or copy change. They are also vibe-coding a tool of their own to answer stakeholder questions, and building an agent where designers paste a prototype link or screenshot and see what past user data says about it. Both are early and imperfect.
Strategic and discovery studies stay human. Their data is abstract, and judging whether something is a real discovery or a generic idea takes a researcher’s head; AI is used there only to search transcripts. “We’re definitely not moving towards full AI-ization of our research,” Gabija says.
AI moderation is used carefully, with user consent, and so far only with panel respondents, not Nord’s own users. A real moderator matters to them: users regularly say how nice it is to be asked, and a human can follow up on surprising moments. “With AI, you don’t have this control. It’s a delicate balance for us.”
Avoiding misleading conclusions
Gabija pushes back on the premise that AI can summarize 100 interviews in a minute. Give it hundreds of pages and it starts sampling. In a couple of studies, the team instead condensed each transcript into a short summary holding only the snippets that answer the research questions, then analyzed those summaries.
The real safeguard is a human in the loop. Ideally the person who moderated the interviews also analyzes them: “Then you know the data by hand, and you know whether AI invented something or is missing something.” Even with AI-moderated studies, the team watches essentially all the recordings and reads every transcript, so it knows the raw data well.
Privacy and customer trust
Nord’s customers choose it for privacy, and that shapes every research tool decision. The team follows standard practice: consent forms, full anonymization, and no personally identifiable information such as names.
Adopting a new tool is deliberately hard. A specialized technical team reviews anything researchers want to try and approves or rejects it. Some popular AI tools are off the table because they don’t meet Nord’s security and privacy standards. “As a research team, this is not our specialty,” Gabija says. “We have a dedicated team that figures this out.”
Testing designs and messaging
Besides the prototype agent, the team runs an AI-powered research repository: a project in an AI tool where any stakeholder can test an idea against existing research data. For new designs, it uses unmoderated testing in Ballpark, an AI-powered tool. An AI wizard helps build the test, AI can moderate it, and AI summarizes qualitative results or drafts a report. A test can now go from launch to data within a day.
Researchers were already comfortable with unmoderated testing, so the real win is democratization. Nord encourages designers and product owners to run research themselves, and AI features remove the “I don’t know where to start” anxiety.
On in-app and website behavior, privacy again sets the limits. Nord prides itself on collecting no user logs, so data analysis teams see only aggregated trends, not individual journeys. AI isn’t much involved there yet. Its role for researchers is translation: turning analysts’ findings into plain language for people who don’t work with data.
A success and a failure
The biggest success was the team’s first AI-moderated study. A stakeholder needed a concept tested, with in-depth data and numbers, in three days; the normal process took two weeks. The team was happy with the interviews and the follow-up questions AI asked. Because AI can run many interviews a day, they also collected quantitative data, and the results fed into important decisions at Nord.
There was no single big failure, Gabija says, just small letdowns where an analysis attempt produced generic, poor-quality answers. The visual side is the hardest. The team recently started vibe-designing prototypes, and copying something that already exists can go wrong in amusing ways, like a person in a mock advertisement who ends up without a proper head.
Her overall stance is pragmatic: use AI to win back time on structured work, and keep people at the center of the research that needs context, judgment and a human voice.
Final Thoughts
Gabija Alčauskienė’s experience at NordVPN offers a practical example of how AI is reshaping UX research workflows. Beyond accelerating analysis and reporting, these tools are making research findings more accessible and creating new opportunities for collaboration between researchers, designers, and product teams.
However, faster research does not automatically mean better research. Recognizing unexpected findings, understanding business context, and building trust with users still require human expertise. As AI becomes more integrated into research workflows, Gabija’s perspective highlights the importance of combining technological efficiency with critical thinking, meaningful user interactions, and informed decision-making.
Connect with Gianluca Lambiase:
