How Enov Uses Recollective to Capture the Insights That Can't Wait

Introduction: A Pioneer in Online Communities

Enov is a French marketing research agency that has been running online communities for more than twelve years. They helped move the category from an experimental methodology to a standard part of the industry's toolkit. The agency operates across multiple markets and industries, including beauty and cosmetics, energy, telecom and automotive, with clients that include L'Oréal, Orange and EDF. Their research practice is built on hybrid methodology: combining asynchronous and synchronous approaches, short-term and long-term communities.

Mariel Turriza Garcia is a Senior Research Executive at Enov who has been with the agency for seven years, using Recollective from day one. In that time she has watched the platform expand from a qual research and insight community tool into one that incorporates AI analysis, automatic translation and now AI-moderated conversational interviews. Her relationship with Recollective is shaped less by vendor loyalty than by a practical standard: the tools she uses are the ones that get her to better results, faster. When something new appears on the platform, Enov tests it, understands it and decides whether it earns a place in their methodology. Conversational tasks — Recollective's AI-moderated interview tool — earned that place quickly.

Background: The Challenge of Being in the Right Place at the Right Time

Enov runs a significant volume of home use test research for beauty and cosmetics brands. In this kind of study, participants receive a product like a moisturiser, a shampoo, a skincare treatment. They’re then asked to use it at home and report on their experience. The research value in these studies is concentrated into a very specific window: the moments immediately after a participant uses the product, when impressions are fresh and reactions are unfiltered. A few hours later, that richness begins to fade. A day later, it's often gone.

The challenge has always been operational. Asynchronous research solves the participant's scheduling problem because they can complete a task whenever it suits them, not when a researcher is available. But this creates a corresponding problem for the moderator: the moment of highest insight value happens on the participant's time, not yours. For Enov, running studies across France, China, the United States and other markets, real-time moderation at that precise moment is impractical. The time zones alone make it nearly impossible to consistently have a researcher available when a participant in Shanghai finishes their morning skincare routine.

The traditional workaround is to review responses manually and follow up later, but "later" is never quite right and the process of individually chasing participants for additional detail doesn't scale. When a community has hundreds of participants across multiple markets, the moderator time required to probe each one meaningfully is substantial. Plus, every hour spent chasing a response is an hour not spent on analysis.

The Solution: AI-Moderated Interviews 

When Recollective introduced AI-moderated Interviews with “Conversational Tasks”, Enov recognised the solution immediately. Rather than waiting for a researcher to be available, the AI moderator is present at the exact moment the participant completes a task — probing, following up and adapting in real time based on how the conversation unfolds. As Mariel puts it: "Conversational tasks are like having an AI assistant that's there when it needs to be, to probe on the exact right information." For home use test research in particular, this was a direct answer to a problem that had existed for years.

The Approach: Using AI-Moderated Interviews in Practice

The way Enov constructs a conversational task follows a logic that mirrors good research practice. Mariel describes their approach as moving from broad objective to specific probe points: "Overall I would like to know this — probe on these points." 

Prompts don't need to be long. What matters is that the AI understands the goal and has enough context to recognise a relevant answer when it encounters one. For a new product, that might mean giving the AI background on who the participants are, what they've been asked to do and what specific aspects of the experience Enov most needs to understand. For a B2B audience, the prompt can instruct the AI to adopt a more formal register. For a younger consumer panel, a more conversational tone. The system adapts accordingly.

Enov started with Recollective's built-in sample prompt templates to understand the underlying logic, then moved to building their own as their confidence grew. The templates were useful for calibration — not as a permanent crutch but as a way of understanding how much detail the AI needs and how to structure instructions clearly. Within a short time, Enov's researchers were prompting from scratch for every project, tailoring objectives and context to each client and audience without needing a template to guide them.

That prompting practice is one part of how Enov works with conversational tasks. They also combine several platform features that change what the AI can do within a conversation.

• The ability to pipe a participant's earlier responses directly into the AI's context is one Mariel highlighted as particularly powerful — it enables the AI to reference what a participant said previously and probe specifically on that, creating the feeling of a real conversation rather than a cold re-start. Previously, that kind of contextual follow-up required a researcher to manually review each participant's response history before writing a personalised probe. Now the platform handles it automatically. 
• Voice input (participants speaking rather than typing their responses) has also changed the quality and volume of what people share, particularly with younger participants more comfortable talking than writing. 
• Finally, when Recollective introduced configurable summary guidelines, allowing researchers to define a consistent output structure across all conversations in a study, it resolved the last significant friction point in working with conversational task data at scale.

Participant Experience: Transparency, Trust and Engagement

One of the first questions Enov had when testing Conversational Tasks was whether participants would accept them. The answer has been clear: they do, provided the research team is transparent from the start. In each study, Enov discloses upfront that participants are interacting with an AI moderator. This is not presented as a limitation; it's simply part of how modern research works and participants who have been involved in online communities for any length of time understand and accept it readily.

Engagement and completion rates have held steady. Enov's concern going into the first live project, that AI moderation might reduce participants' willingness to share in depth, did not materialise. "We've seen completion of tasks that hadn't changed from when we were moderating it ourselves," Mariel notes. "People give just as much detail. They're just as engaged as they are when it's a real person behind the keyboard." A typical Conversation Task with two or three probe points runs for two to five minutes, which is not materially different from a standard asynchronous activity. The AI also handles disengaged or uncooperative participants sensibly, recognising when a conversation has reached its natural limit and moving on rather than continuing to press for answers that aren't coming.

The addition of voice input has been a meaningful change to participant experience. Typing responses to research questions requires a certain level of deliberate effort; speaking them does not, and the answers that emerge are often richer and more spontaneous as a result. For Enov's multi-market research, this matters particularly in communities where written response rates have historically been lower.

Impact on Research Teams: From Manual Chasing to Strategic Analysis

Conversation Tasks eliminated the most time-consuming part of async moderation for Enov, enabling their team to move faster. Manual probing, including reviewing each participant's response, identifying what's missing, writing a follow-up question and waiting for a reply, has been substantially reduced. The AI handles first-pass probing at the moment it matters most. Researchers come in after the fact to review, fill any remaining gaps and move to analysis.

The reduction in mental load is real. Moderators are no longer tracking, across dozens or hundreds of participants, who has been followed up with and who hasn't. The AI reliably and consistently takes on that responsibility, without the lag inherent in human-led asynchronous moderation. The time saved flows directly into analysis; building conclusions, developing recommendations and constructing the deliverable that actually reaches the client. "We've been able to get insights and information that otherwise would've had to wait hours or days before we could collect them," Mariel says. "It's definitely allowed us to move faster, so we can spend more time on the analysis."

Importantly, Conversation Tasks have not changed the structure of Enov's research teams or the role their community managers play. Enov made a deliberate decision not to let the feature replace other task types. Discussion forums, live video sessions and standard asynchronous activities all remain part of their methodology. The AI moderator handles a specific function: probing at the moment of highest relevance. Everything else continues as it did before, with experienced community managers and moderators at the centre of how each community runs.

Results: Scalable Qual at a Level Previously Out of Reach

The outcome Enov mentions most consistently is scale. Conversational tasks have allowed Enov to run qualitative research with participant volumes that would previously have required a proportionally larger team to moderate meaningfully. "It's been a bit of a game changer when it comes to bigger-scale qual," she says. "It's allowed us to interrogate a lot more people than we would have a few years back." 

Multi-market execution has become more manageable. The same study logic runs across markets in multiple languages, with Recollective's translation infrastructure supporting what would otherwise require separate moderator deployment in each region. A study running simultaneously in France, China and the United States no longer requires a researcher available across three time zones — the AI is present whenever the participant is ready.

“We've had conversations about AI with our clients for years now — it didn't start this year or yesterday. I think there was no apprehension in that regard. Quite the opposite, actually. It's been more like an opportunity. We've tried to approach it, both at Enov and with our clients, as an opportunity to work better, work faster, be more relevant and of course be more value-driven partners."

Enov's clients have not been apprehensive about the shift. AI has been part of the agency's research practice and client conversations since Recollective first introduced it and the introduction of AI-moderated interviews was treated as an extension of existing capability rather than a departure from established methods. Clients are not receiving different insights — they're receiving insights that are more timely and more contextually grounded, captured at the moment that matters rather than reconstructed hours later. Enov's position is deliberate on this point: conversational tasks make insights more relevant, not just more convenient.