Dentsu-owned media agency Carat wants to understand audiences better, and its solution is an introvert’s dream: cut down on human conversations and bring in the bots.
On Thursday, Carat announced a partnership with Vurvey Labs (a portmanteau of “video” and “survey”), an agentic AI-powered research platform that uses short video surveys to quickly get insights about people without having to do traditional, time-consuming interviews.
Vurvey allows Carat to track trends and interests within specific demographics by pulling first- and third-party data from multiple sources, including social media platforms, a brand’s own internal data and Dentsu’s past campaign results.
From there, Vurvey’s AI agents are trained to understand a specific audience and the best way to reach them at the current moment.
By the people, for the people
Vurvey markets its tools as “people-powered AI” – a phrase that, at first glance, sounds awfully oxymoronic.
But Vurvey’s agentic AI isn’t a traditional large language model (LLM), Michael Liu, EVP and head of innovation at Carat, told AdExchanger.
Unlike a traditional chatbot, it’s constantly evolving to reflect real people and day-to-day variations in trends.
A lot of agentic AI is “primarily instruction-based” and funnels data that’s already on the internet to an LLM, said Chad Reynolds, founder and CEO of Vurvey Labs. Vurvey’s agents, on the other hand, are trained on real people to fit a specific consumer profile.
Carat determines audience preferences by sending out Vurvey’s namesake video surveys (i.e., “Vurveys”). They’re sent out in the form of URLs that link to a list of questions about a customer’s habits and preferences. The customer is instructed to record their answers in the form of a 20-minute or so video, which is “condensed into data to enrich these consumer profile agents,” said Liu.
One of the benefits of a video survey, he added, is the immediacy of the insights you can gather. AI can synthesize hundreds of interviews in a fraction of the time it would have taken a person to conduct and analyze the information.
AdExchanger Daily
Get our editors’ roundup delivered to your inbox every weekday.
Daily Roundup
Within a brand, Carat designs different agents that represent different audience segments in order to reflect variation in spending habits and brand loyalty.
Vurvey has also built what Liu describes as a “trend-hunter agent” that scours the internet for “what’s popping” (yes, direct quote), and then determines which of these “popping” trends are relevant to the audience that a given agent represents.
For a casual dining client, for instance, Carat built two personas: one for loyalists (people who eat at one of the client’s restaurants at least monthly) and one for prospects (those who visit the brand’s competitors, but not this client’s locations).
Carat built a persona based on detailed information about the brand to reflect its “characteristics and values,” said Liu. By accessing consumer personas, Carat was able to design an interactive AI agent that can mimic different segments of the brand’s audience, better understanding each segment’s preferences and what differentiates it from others.
The heart of it
Carat is hopeful that its partnership with Vurvey will reinforce the idea that “AI is only powerful and only valuable when you power it with real people,” Liu said.
Vurvey’s model brings culture to the forefront of the conversation, giving Carat access to more personal insights into its audience.
The powered-by-people model “codifies” the personality traits, psychographics and behaviors that Vurvey has compiled from millions of interviews, said Reynolds, and is constantly evolving so that its agents are “truly reflective of today, not six months ago.”
One of the reasons that Carat chose to work with Vurvey, said Liu, is rooted in a shared belief that despite the synthetic nature of AI, it can help companies see their audiences as real people and understand them on a more personal level.
And that’s exactly what Vurvey set out to do.
“We built the people model to really dig into the details of who someone is,” said Reynolds. When agents are built on specific insights from real people, he added, they “come to life with reflections into the real world.”
“Come to life” might be a bit of a stretch (at least, here’s hoping), but it’s definitely a whole new layer of insight into – and reflection of – human behavior.
