What Does an AI Research Strategy Actually Look Like for Insights Teams?

By Adam Mertz, Chief Strategy Officer, Discuss Last updated: May 2026

Key Takeaways: A real AI research strategy treats every study as a building block, not a finished deliverable. At HelloFresh, that means weekly in-home ethnography across ten countries, with 30 to 40 people from across the business watching live recap reels every Friday. The infrastructure behind that cadence is what a working AI research strategy looks like.

A few weeks ago, our team was in Frankfurt for Succeet, one of the leading market research conferences on the European circuit. I shared the stage with Jo Lindenberg, Director of Global Consumer Insights at HelloFresh. We knew the topic would land. We didn’t expect people to be turned away at the door due to fire code limits.

That told me something. The conversation about what a genuinely modern AI research strategy looks like is one the industry has been waiting to have out loud.

How do insights teams scale qualitative research without losing depth?

The strategy Jo has built at HelloFresh is deceptively simple: treat every study as a building block. Every insight stays alive and queryable. The knowledge compounds.

HelloFresh tracks sentiment across 18 markets and ships nearly a billion meals a year. Traditional qual, scoped, fielded, and delivered on a project timeline, can’t keep up with the questions that arrive every Monday morning. Jo’s frame for this was precise: “Every week, we see changes in our tracking data. And every week, we need to know why.”

That’s an organizational design problem. Most insights teams are still solving it with tools and timelines built for a different era. The result is familiar: research that dies in slide decks, redundant studies because no one can find the last one, and a stakeholder somewhere making a decision on instinct because the insight arrived three weeks too late.

Before adopting AI-moderated approaches, HelloFresh ran in-home ethnography the way most global teams do: a small number of participants, limited geographic coverage, significant coordination across scheduling, filming, and production. The output was valuable. The process wasn’t built for frequency or scale.

Moving to AI-moderated interviews on Discuss changed that equation, while keeping the research human-centered.

Now they run in-home ethnography every single week, across ten countries, with 18 respondents, and 30 to 40 people from across the HelloFresh business watching live recap reels together. Not as a one-time project. Every week, on rotation.

Why do consumers open up more with AI moderation?

Jo had been a skeptic going in, her words, and she’s still discerning about where AI earns its place. But the quality of what came back from AI-moderated sessions caught her off guard.

Consumers were candid. Uninhibited. Often more open than they’d be with a human moderator.

Her read on why: COVID normalized video conversations, and ChatGPT normalized talking to AI. Add what she called the “stranger on the bus” effect. No one to perform for, no judgment, and the view inside people’s homes is actually more authentic with AI moderation.

Because respondents complete sessions at their own pace, the footage is genuinely UGC-style: a messy kitchen, their prized air fryer, their own perspective. As Jo put it: “They’re the ones holding the camera.”

That shift in perspective changes what becomes visible.

What does an AI research strategy look like when it’s fully operational?

The weekly cadence of AI interviews is only one layer. The more consequential shift is what HelloFresh is building on top of it.

Their team is constructing a centralized intelligence layer by structuring years of qualitative data in a way that can be accessed dynamically. Using Discuss Virtual Personas, they are creating AI representations of customer segments grounded entirely in real interview data.

A brand manager exploring messaging doesn’t need to initiate a new project to get directional feedback. They can engage with a persona informed by thousands of prior conversations and develop a more informed point of view in hours rather than weeks.

Jo described where they are now as “spoiled for data.” The real challenge, she said, is no longer generating insights. It’s giving everyone in the business fast, direct access to what already exists.

That’s the shift: from research as a reporting function to research as a living asset that compounds with every study you run. Every AI interview, every live IDI, every data point that feeds a Virtual Persona makes the next question cheaper and faster to answer. At HelloFresh, that’s already operational.

Forrester recognized this direction when they named Discuss a Leader in the Q1 2026 Experience Research Platforms Wave, specifically for the AI capabilities that let teams scale qualitative research globally without sacrificing depth or rigor. The category is maturing fast, and the distance between organizations that have built this kind of intelligence layer and those still treating research as a series of isolated projects is widening.

How do you know if your organization is ready for this kind of AI research strategy?

Jo embodies what the AI-empowered researcher looks like: someone who used AI to do work that wasn’t previously possible, at a pace and scale that changed how her entire organization relates to its customers. That’s the role insights can play when the infrastructure is built for it.

The next time a stakeholder asks you an insight question, ask yourself how much of the answer already exists somewhere in your organization, buried in a transcript, locked in a dashboard, sitting in a slide deck from Q3.

That answer doesn’t have to stay buried.

Frequently Asked Questions

What is an AI research strategy for insights teams? An AI research strategy treats qualitative data as a continuously growing asset rather than a series of closed projects. It combines AI-moderated interviews for scale and frequency with a centralized intelligence layer, such as AI-powered customer personas grounded in real interview data, so that insights remain accessible and queryable across the organization.

How does AI moderation affect the quality of qualitative research? When designed well, AI moderation can produce more candid responses than human-led sessions. Respondents tend to be less guarded without a human moderator present. The tradeoff requires careful design: AI moderation works best when it’s grounded in real consumer data and reviewed by skilled researchers who can interpret what they’re seeing.

How do you scale qualitative research across multiple countries from one platform? Platforms like Discuss run AI-moderated in-home ethnography simultaneously across multiple markets, with consistent methodology and centralized analysis. HelloFresh runs weekly sessions across ten countries with 18 respondents per cycle, with recap reels distributed to 30 to 40 stakeholders the same week.

What is a Virtual Persona in qualitative research? A Virtual Persona is an AI representation of a customer segment built entirely from real interview data. Unlike synthetic respondents generated from public data, Virtual Personas are grounded in an organization’s own research history. Teams can query them directly to get directional feedback without initiating a new study.

How is Discuss positioned in the qualitative research market? Discuss was named a Leader in the Forrester Wave for Experience Research Platforms, Q1 2026, specifically for AI capabilities that support global qualitative research at scale. Discuss has processed over 10 billion tokens of unstructured consumer research data and is one of only 150 organizations globally recognized by OpenAI for data processing volume, and the only market research company on that list.Check out the Succeet session recording here. Download a complimentary copy of the Forrester Wave™: Experience Research Platforms, Q1 2026 here. And if you want to see how HelloFresh’s approach translates to your own research stack, let’s talk.approach translates to your own research stack, let’s talk.