I've been doing UX research for 25 years. In 1999, I founded one of Germany's first user-research-based strategy consultancies. I've witnessed the shift from paper prototypes to Figma, from focus groups to remote testing, from handwritten notes to automatic transcription.
What's happening now feels different.
I'm building a tool that automates in-depth interviews. At the same time, I know that in-depth interviews are the crown jewel of qualitative research – they require experience, empathy, sensitivity, curiosity, and the courage to ask uncomfortable questions. To people you've never met.
That sounds like a contradiction. It isn't.
The State of the Industry
Before we talk about agentic AI, we need to talk about the industry. Because the numbers on AI adoption only tell half the story.
UX research is in crisis. According to UXPA surveys, 35% of organizations have lost staff, 37% have had layoffs. The Nielsen Norman Group writes: "A year ago, UX felt like it was on trial." The job market is slowly stabilizing, but uncertainty remains.
Into this climate drops the news: 78% of researchers believe, according to Qualtrics' 2026 Market Research Trends Report, that AI agents will handle more than half of all projects end-to-end within three years. 15% are already using agentic AI today.
You have to read this number in context. Behind it lies not just technology optimism, but also fear. Fear for jobs, economic uncertainty, the worry: What does AI mean for me, my profession, my role?
The Nielsen Norman Group calls 2026 the "year of AI fatigue." People are tired – tired of being told they'll be replaced if they don't "vibe code." Tired of tools that don't fit into real workflows. Tired of explaining why automated decisions are risky.
What "Agentic" Means
An agentic system differs from classic AI tools in four ways.
Goal orientation: The system receives a goal, not an instruction. "Find out why users abandon checkout" instead of "Transcribe these five interviews."
Planning: It breaks down the goal into steps on its own.
Execution: It carries out the steps without someone confirming every click.
Adaptation: It responds to what happens. If a question doesn't work, it rephrases.
The difference from an assistant: An assistant waits for instructions. An agent acts autonomously.
What's Gained
The question isn't what's lost when AI conducts interviews. The question is what's gained.
In-depth interviews aren't just the crown jewel – they're also time-consuming and labor-intensive at every phase. Recruitment, scheduling, conducting, transcription, analysis. Even customer-centric companies, for whom qualitative user research was standard, are cutting budgets. They're leaning harder into data-driven decisions because they no longer have the time or money. The pressure is high, release cycles are getting faster.
The necessary research often simply isn't being done.
That's the real problem. Not that AI conducts interviews, but that interviews aren't happening at all. Good products emerge from dialogue with people. But that dialogue isn't happening. Instead, teams interpret data, develop hypotheses, and make gut decisions.
AI can make that dialogue possible again – for companies that otherwise couldn't afford it.
What Can't Be Replaced
Of course there are situations AI can't handle. Asking questions from intuition, from experience, from intrinsic curiosity. Recognizing the moment when someone says one thing but means another. The pause that reveals more than any answer.
Communicating results to stakeholders not just to inform them, but to move them toward a shared and binding perspective – no AI can do that.
But much has changed by 2026. Context windows have grown larger, agentic capabilities stronger, data interpretation and analysis better. AI keeps evolving. The Nielsen Norman Group writes: "We will see core AI technologies incrementally improve their 'jagged' capabilities, potentially reaching watershed moments for user-research activities."
The boundaries are shifting. That doesn't mean they're disappearing.
The Trust Problem
The Nielsen Norman Group identifies trust as the biggest UX problem for AI experiences in 2026. People who've been disappointed by AI features adopt new systems more hesitantly.
For research, this means: An agent that gets an interview wrong doesn't just damage that one study. It damages participants' trust in AI-assisted research overall.
Agents are often launched before they're ready. The result: bad experiences, growing skepticism. Every bad interaction makes the next adoption harder.
How QUALLEE Implements This
QUALLEE is an agentic research tool. The AI interviewer conducts conversations independently, asks follow-up questions based on responses, extracts themes automatically. But we've built in deliberate limits.
Humans define the goals. The agent executes, but research questions come from the team.
Raw data stays accessible. QUALLEE doesn't deliver finished summaries you have to accept – it delivers quotes, the raw voice of users. The researcher interprets.
Participants know they're talking to AI. That's not just EU AI Act compliance – it's respect.
For complex questions, there are hybrid projects. AI for volume, human expertise for depth.
What Remains
UX research will become more important again. The craft will matter more, the intimacy and authenticity that emerges from dialogue between people. The human element itself will be valued more as AI becomes more prevalent.
But that won't change how money and time get allocated. There will be more data, more computing power, more intelligence at the push of a button. AI-generated simulations will be incredibly good.
But the money made from people still flows from actually talking to those people.
Frequently Asked Questions
What is Agentic AI in the context of UX research?
Agentic AI refers to AI systems that independently execute multi-step workflows. In UX research, this means: Instead of automating single tasks like transcription, agents handle entire processes – from conducting interviews to extracting themes.
Does Agentic AI replace human UX researchers?
No, but it changes the role. Researchers become strategists who define goals and interpret results. Agents increasingly handle the operational execution. This requires different skills, but not less expertise.
How reliable are AI agents for qualitative research?
For standardized use cases with clear goals, they work well. For exploratory research, sensitive topics, or situations requiring empathy, humans remain superior. The art lies in proper deployment: agents for volume, humans for depth.
What does the EU AI Act mean for agentic research tools?
From August 2026, full transparency and documentation requirements apply. Users must know they're interacting with AI. Providers must classify and document risks.
Try It Yourself
Experience what an AI-led interview feels like. In our current research project, we're exploring how people interact with AI in their daily lives. The conversation takes about 20 minutes.
