AI reshaping product research, but trust remains key

Lyssna has published survey findings showing that 94% of product professionals use AI in their research workflow. The results are based on responses from 428 attendees surveyed at Future Product Days in Copenhagen.

The research points to broad adoption of AI tools among designers, product managers, engineers and researchers. It also shows that concerns over privacy and accuracy remain a brake on wider use with research data. More than half of respondents, 55%, said they use AI throughout their research-to-decision workflow, while 62% said an AI assistant such as Claude or ChatGPT is part of their process.

The findings suggest AI is changing how research is carried out inside product teams rather than replacing the work altogether. Nearly half of those surveyed, 47%, said AI has led them to do more research, while 20% said they now rely on AI more than on doing the research themselves.

Research activity also appears to be spreading beyond dedicated research roles. Designers made up 46% of respondents and product managers 28%, while only 4% identified as researchers. That distribution suggests research tasks are increasingly being handled by broader product teams.

Tool connections

Almost half of respondents, 49%, said they had connected an AI assistant directly to their research or product tools and used it in their regular workflow. The survey cited examples including connections made through the Model Context Protocol.

Engineers were the most likely group to have taken that step, with 72% saying they had connected an AI assistant to their tools. The figure was 55% for product managers and 39% for designers. Some respondents said company policy prevented them from linking AI tools directly to internal systems.

Those who had integrated AI assistants into their workflow were also more likely to say the technology was increasing their research activity. Among regular users of connected assistants, 56% said they were doing more research, compared with 24% of people unfamiliar with connecting AI to their tools.

Post-study strain

The survey found that the hardest part of the process often comes after a study ends. Manual synthesis or write-up was named as the biggest slowdown by 32% of respondents, followed by pulling together data from multiple tools at 26% and getting the right people into the room at 19%.

Respondents also pointed to a newer burden created by AI adoption itself: checking AI-generated output for errors and bias. That concern feeds into the wider issue of trust, which emerged as a central theme across the findings.

When asked which AI workflow would most improve their research process, respondents ranked auto-synthesising findings across studies first at 33%. Auto-generating reports from raw data followed at 28%, suggesting teams see the greatest potential in reducing the manual work involved in turning research inputs into decisions.

The appetite for more direct outputs was also clear elsewhere in the results. Four in five respondents, 81%, said they would value an AI-generated, decision-ready report based on their research data.

Trust issues

Despite the high level of adoption, 27% of respondents cited data privacy as the biggest barrier to using AI directly with research data, and 25% pointed to trust in accuracy. While 29% said nothing was holding them back, the results suggest confidence in how AI handles sensitive information, and whether it produces reliable conclusions, remains unsettled.

The issue was especially pronounced among people who had already connected an AI assistant to their tools but did not use it much. In that group, trust in accuracy was the most common barrier, cited by 40% of respondents.

An open question in the survey drew 258 responses on what people would want an AI assistant to do with their research data. The most common request was help spotting patterns they might miss on their own. Others said they wanted AI to act less like a passive helper and more like a critical partner in interpreting evidence.

One respondent captured that view in a short answer included in the findings: "Challenge me instead of agreeing with everything."

Lyssna said the survey reflects an AI-engaged product audience rather than all product teams because attendees were self-selected and there was no screener. It also noted that researchers were a small part of the sample, with 18 people in that category, meaning researcher-specific findings should be treated as directional rather than conclusive.

"AI hasn't replaced research - it's changed who's doing it," said Tristan Gamilis, Chief Product Officer and Co-Founder at Lyssna.

"Designers and product managers are running more research than ever, and AI is helping them do it. The challenge now is trust. Teams need to know that what their AI tells them comes from real people," Gamilis said.

"The teams that get the most from AI will be the ones that keep it grounded in what real people actually said and did," said Mateja Milosavljevic, Chief Executive Officer and Co-Founder at Lyssna. "That's how you move fast and still make decisions you can stand behind."