How research leaders are scaling qualitative research with AI, and what Forrester found when they looked under the hood

By Adam Mertz, Chief Strategy Officer, Discuss

Discuss is an AI-powered market research platform named a Leader in the Forrester Wave™: Experience Research Platforms, Q1 2026 — the only unified qual, quant, and AI solution in the market. We build AI agents, AI-moderated interviews, Virtual Personas, and continuous intelligence infrastructure so research teams can ask anything and answer everything, without starting a new project every time. Last week, we hosted a live webinar with Senem Guler-Biyikli, the Forrester analyst who authored that Wave report. What she shared is worth putting on the record.

Senem has spent months this year interviewing research and design leaders about how AI is actually changing their workflows — not how vendors say it’s changing them, but how the people doing the work describe it. Her findings confirmed some things we believe strongly at Discuss, challenged a few assumptions in the market, and gave me a sharper frame for where this industry is going.

Who’s actually using qualitative research platforms now and why it matters

Senem’s Forrester Wave evaluation covered eight platforms and involved vendor questionnaires, live product demos, strategy briefings, and direct interviews with reference customers. Before getting into AI workflow findings, she walked through four patterns that emerged from that process.

First, Experience Research Platforms now serve a much broader range of roles than they did a decade ago. 

Where these tools were once the domain of professional researchers alone, they’re now being used by designers, product managers, marketers, and sales professionals or basically anyone who needs to understand customers to make better decisions. That shift creates a real design challenge for platforms. Supporting multi-method research across a wide range of skill levels and use cases in one coherent product is hard. The vendors that solve it gain a structural advantage.

Second, there’s also a sharp split between organizations with high AI trust and those with low AI trust, and it’s showing up directly in how teams use research platforms.

Early adopters have moved past basic AI summaries — they expect to have conversations with their data, to probe and follow up and interrogate, the way you would with a knowledgeable colleague. Organizations with strict AI governance policies, on the other hand, often can’t take advantage of built-in AI features at all. Both groups exist inside the same buyer market, which means platforms need to be built for both.

Third, how effectively a platform uses AI has become a primary purchase criterion. 

Every platform in the Forrester Wave evaluation uses AI. What varies significantly is where in the research lifecycle it’s applied — some focus on analysis and synthesis, others embed AI earlier in planning and study design — and whether customers actually find those features useful. The gap between “we have AI” and “our AI works” is where buying decisions are getting made.

The fourth finding surprised Senem enough that she called it out explicitly: AI moderation is generating genuine enthusiasm from buyers.

Synthetic audiences and other AI-generated data tend to produce interest mixed with caution. AI moderation is different. The customers she spoke with were actively hoping the technology would mature quickly because they saw a clear path to doing qualitative research at a scale that wasn’t previously possible. That enthusiasm is something we’ve seen reflected in our own customer conversations — and it’s shaping how we think about the roadmap.

The risk hiding inside AI-led research workflows

Senem opened with a finding that deserves more attention than it typically gets: Forrester predicts that this year, at least two major scandals will result from firms acting on AI-led customer research.

That’s a specific, confident prediction from the top analyst firm covering this space. The mechanism she described is straightforward: overstretched teams hand off planning and execution to AI agents, overestimate AI accuracy, and underestimate what research expertise actually contributes. They get results. They act on them. The results were wrong.

The data behind that prediction is striking. According to Forrester’s survey of AI decision makers — people deeply involved in their organization’s AI strategy — 69% believe GenAI tools will always produce the same output given the same prompt. 83% believe AI models are good at looking up and validating facts. 78% consider AI outputs trustworthy. These beliefs are simply incorrect. GenAI is probabilistic, not deterministic. It confabulates. It reflects the biases in its training data. When those misconceptions get embedded into a research workflow, the outputs look credible while being wrong.

This matters for market research specifically because the output of research — unlike, say, a poorly generated image — is designed to inform decisions. Bad research that looks good is more dangerous than obviously bad research.

What happens to the researcher role as AI takes on more of the work

Senem was direct about this: she’s optimistic about the researcher role. So am I. But optimism shouldn’t be confused with “nothing changes.”

What’s changing is where the work lives. As AI takes on execution-heavy tasks — drafting discussion guides, generating screeners, coding transcripts, building prototypes — the researcher’s role shifts toward judgment, critique, and decision-making. The question moves from “can I do this?” to “should we do this, and how do we know if the result is right?”

This matters for how research teams hire and how they position themselves internally. The researchers who will thrive are the ones who can answer: what biases might this AI tool introduce? Is AI moderation appropriate for this study design, or will it miss something a human moderator would catch? What are the trade-offs of using a Virtual Persona for early directional signal versus fielding with real participants? These are not questions that come bundled with AI tools. They come from research expertise.

Senem pointed to a useful data point: design teams that plan to grow intend to hire more researchers. The concern that AI makes researchers redundant isn’t what the data shows. The concern worth having is whether researchers are positioning themselves as the people who can answer those harder questions.

How to scale qualitative research with AI moderation

Of all the findings from Senem’s customer interviews, AI moderation generated the most enthusiasm. That surprised her, and she said so directly. Synthetic audiences and synthetic data tend to generate interest mixed with hesitancy. AI moderation is different. Buyers were actively hoping it would succeed so they could scale qualitative research in ways that weren’t previously feasible.

At Discuss, we’ve watched this play out with customers in real time. HelloFresh went from occasional in-home ethnography — expensive, logistically heavy, limited to a handful of respondents — to running AI-moderated consumer closeness sessions every week, across 10 countries, with 30 to 40 internal stakeholders observing. The volume of insight didn’t just increase. The organizational relationship to consumer understanding changed.

That’s the practical answer to what scaling qualitative research with AI moderation actually looks like: not just more respondents, but more teams inside the organization staying connected to real consumer voices, continuously, without the logistics overhead that used to make that impossible.

Senem made an important distinction that often gets lost in the AI moderation conversation: it’s not a replacement for in-depth human research. For CX teams that are survey-heavy, AI moderation opens a door to qualitative feedback they couldn’t previously access at scale. For research questions that require genuine depth, human-led sessions still belong in the workflow. The sophistication is in knowing which is which, and that judgment belongs to the researcher.

What “always-on” qualitative research actually looks like in practice

The second half of our conversation with Senem moved to where this all leads: research as continuous intelligence rather than a sequence of projects.

Forrester’s State of Design survey found that 52% of design professionals use centralized insights to train their AI tools. That’s a signal about where research repositories are heading, not just as archives, but as the training substrate for internal AI. The teams that have been building structured, accessible research data for years are sitting on a significant advantage. The teams whose research lives in presentation decks are at a real disadvantage.

The shift Discuss is building toward is exactly this: every study should make the next one smarter. Virtual Personas built from real customer interview data — not open-internet AI — let anyone in the organization ask questions of their target audience in seconds, informed by years of actual human responses. For example, Mondelez used their Discuss Virtual Persona of a 50-plus aged premium chocolate buyer to answer questions for a leadership meeting the same day they were needed, without running a new study. That’s what it looks like when research stops expiring in slide decks.

The broader intelligence layer connects qualitative data, quant surveys, and past research into a single queryable system where any team member can ask a specific question and get an answer that pulls from the full picture. A question gets answered. Not a new two-week project.

Why the best qualitative research platforms combine AI moderation with human-led research

At Discuss, we talk about keeping humans in the loop, and it’s worth being precise about what that means, because it’s easy to misread as skepticism about AI. It isn’t.

Discuss is an AI-first company. We have built AI agents that run end-to-end research workflows, AI-moderated interviews, and Virtual Personas trained on validated customer data. We were named a Leader in the Forrester Wave for Experience Research Platforms this year specifically for the depth and effectiveness of those AI capabilities. The human-in-the-loop position comes from knowing exactly how AI works and where it needs researcher expertise to produce results that are actually trustworthy.

Senem’s prediction about research scandals this year might sound like an argument against agentic research or AI moderation. But it’s not. It’s an argument against using public general-purpose models without understanding their limitations, handing research planning entirely to AI without oversight, and treating probabilistic outputs as ground truth. The teams that get the most from AI are the ones who stay at the center of it: orchestrating, questioning, validating, and making the call on when AI is the right tool and when a human moderator is the right tool.

That’s the distinction Discuss is built around: AI that makes researchers more capable, not AI that makes researchers optional.

About Discuss. Discuss is an AI-powered market research platform and the only unified qual, quant, and AI solution in the market. Discuss offers AI Agents for end-to-end research workflows, AI-moderated interviews, Virtual Personas built from validated customer data, and a continuous intelligence layer that makes every study searchable and queryable. Named a Leader in the Forrester Wave™: Experience Research Platforms, Q1 2026. Used by global organizations including HelloFresh, Mondelez, Meta, P&G, and Nike. Learn more at discuss.io.

Download a complimentary copy of the Forrester Wave™: Experience Research Platforms, Q1 2026 here. To see how Discuss helps research teams build a continuous intelligence infrastructure, request a conversation.