In this article we're going to look at how we have been eintegrating AI into our work practices. We have begun using AI to mimic focus group feedback, especially for testing creative ideas, messages or visual content before formal launch. This addresses a common client question: “How will our target audience react?” – but does so in a fraction of the time of a traditional focus group.
Our team developed a process to create virtual personas (detailed audience archetypes) and then have AI “think” and respond as those personas 5. We call this an AI-assisted focus group or synthetic focus group. Here’s how it works in practice:
- Define Personas: Drawing on research (demographics, psychographics, past consumer research), we craft profiles for each key audience segment. For example, for a recent project in the food sector we had personas like “Health-Conscious Young Professional” and “Traditionalist Parent”, each with attributes (age, values, what media they consume, etc.). These were initially written by our strategists and refined with the help of AI to fill in any traits we might overlook (using ChatGPT to suggest realistic details given the demographic).
- Feed Supporting Data: To ground these personas in reality, we leverage our existing data – e.g. YouGov survey findings, past focus group transcripts, social listening data – and embed that into the AI’s knowledge. In technical terms, we use a retrieval-augmented generation (RAG) approach: all our research reports, trend data, etc. are indexed in a vector database that the AI can query. So when the AI persona is asked for an opinion, it will first retrieve a few real facts or insights relevant to the topic (for instance, if the content is an ad about sustainable farming, the AI might fetch data on how audiences feel about sustainability in food). This ensures the AI’s responses aren’t just plausible-sounding fiction, but are anchored in real-world evidence and audience insight.
- Simulate the Discussion: We then prompt the AI to “step into the shoes” of each persona and react to the creative content. For example, Persona A is shown a draft tagline or a storyboard and asked, “What do you think about this? Would it catch your attention? Do you find anything off-putting or appealing?” The AI, drawing on the persona profile and the research data (via RAG), generates a response as that persona 5. We do this for each persona, and even simulate dialogue – e.g. Persona B might respond and Persona A can rebut – essentially an AI-moderated focus group conversation.
- Aggregate Feedback: The outputs are collected into a “report” for our strategists to analyse. We often get rich qualitative feedback: e.g. “As a busy parent, this ad’s eco-friendly message is nice but the benefit isn’t clear – I worry it’s all buzzwords” (from the Traditionalist Parent persona) or “I love the modern design, it feels aligned with my values” (from the Young Professional persona). We compile key themes and even ask the AI to score each concept on metrics like perceived relevance or likelihood to share, if appropriate.
This method has been pilot-tested in-house and with a few clients. For instance, in May 2025, a client in the agriculture sector was keen to test campaign ideas about lamb recipes targeted to different consumer types. Instead of organizing physical focus groups in multiple cities, we ran an AI focus group simulation with four personas corresponding to their audience segments. We provided those personas with the campaign’s visuals and key messages. Within a day, we had detailed “feedback” from each persona.
Food Campaign Synthetic Focus Group. Our client, a national food marketing board, wanted quick insight on which of two creative directions for a promotion would resonate more. We created four personas (ranging from a young fitness enthusiast interested in protein, to an older traditional cook focused on family meals) based on the client’s target demographics. Using our synthetic focus group setup, the AI personas reviewed the campaign ideas. The AI feedback revealed that the modern, health-oriented campaign angle was well-received by the younger persona (“This fits my meal-prep lifestyle”), but left the traditional persona cold (“What’s the story here? I care about family, not just health metrics”). Meanwhile, the more classic campaign concept evoked a warm response from older personas (they loved the nostalgic family dinner scene) but seemed “boring” to the younger ones. These nuanced reactions emerged overnight. Armed with this insight, our team recommended blending the two approaches: we added a family-oriented narrative to the modern campaign concept. The client loved this solution. Not only did we deliver actionable insight exceptionally fast, we saved on research costs – no recruitment or incentive payments needed – and could iterate creatives in an agile way.
The outcome was a campaign that struck a chord across audience segments, confirmed later by strong social media engagement once it launched (which aligned with the AI personas’ predictions). While this AI-simulated focus group isn’t a perfect crystal ball, it gives us a valuable early read that shaped creative development.
Benefits: The primary advantage is speed and agility. Traditional focus groups might take 3-4 weeks to recruit participants, conduct sessions in multiple locations, and produce transcripts & analysis. In contrast, an AI focus group can be set up in a couple of days and deliver initial directional feedback within hours. This means we can test multiple ideas rapidly. For example, we can run 10 different taglines through the AI personas and quickly narrow down to the top two based on simulated reactions. It’s also cost-efficient: aside from our team’s time to set it up, there are minimal incremental costs (no venue hire, travel, or participant fees). We often describe it as “wind tunnel testing” creative ideas – it’s not the same as real world, but it’s an excellent experimental environment to weed out obvious issues and improve concepts early.
So far, our clients have been enthusiastic about this innovation. It’s a novel experience for them to “hear” their customers talk via an AI. One client remarked that the AI-generated persona comments were “eerily similar to what our customers say – it’s as if you read our last focus group report and brought those people to life again”. That’s exactly our aim: use AI to amplify the voice of the customer by digesting all the research out there and expressing it in a vivid, conversational way.