Vurvey Labs’ Jeremy Smallwood on the human element of AI

The AI space is overwhelmingly crowded. Vurvey Labs’ differentiator? Humans.

The company uses video surveys from people (hence the name Vurvey, short for video survey) to create AI personas for brands. Brands, in turn, can work with these virtual consumers — informed by real ones — to develop products, test messaging and uncover needs.

Four years into its venture, Vurvey Labs works with powerhouse brands like Unilever, JPMorgan Chase, Adidas, Vans and Kenvue.

Jeremy Smallwood, the company’s senior vice president of partnership experience, recently spoke with The Current Editor-in-Chief Stephanie Paterik about how brands can scale the focus group, address privacy and overcome AI fears.

What are people not talking about in the AI startup space?

Ironically, humans, to be honest. That loop of the human starting everything off, being part of it through the whole process. I mean, we often talk about what AI can do for us, but really, what can we do for the AI? How can we help it be better? How can we help it solve our problems in more interesting ways? We have to be part of that equation.

How would you describe the AI opportunity for brands?

I think of clients of ours like Lowe’s, who literally built an entire population of 55-year-old contractors. It’s a tough population to reach. So building those personas out in a way that they can iterate and co-create with their audience quickly and do that in a way that gets to bigger ideas.

Kenvue is another client of ours, [it] actually built personas around people with different skin types so [it] can test out new ideas and innovate new products. So again, keeping that consumer close at hand and actually co-creating with them along the whole way.

How did Vurvey Labs launch? What’s the big bet that you’re making?

Vurvey, video survey, just throw them together. Truly, Vurvey started off as recognizing the bottleneck in research.

You spend all this time and effort creating and researching this one segment or this part of the audience, and then you kind of step away from it. To be honest, you sort of lose the plot down the line. So how can we keep that audience with us the whole way.

Break this down from a technological standpoint: How are you building AI personas? How do you take that human element and bring it into this agentic AI space?

We send out our vurveys to our respondents, to your respondents. We’re agnostic, so we’ll send them out to whoever you know you want them to reach.

Those video surveys come back and give us a real inkling about how you’re talking about yourself and who you are. And using our people model, that’s 19 trillion facet combinations. That’s a big number.

If you think about it, you’re not just a consumer [who] walked in to grab a Starbucks today, you’re someone who maybe is from the Midwest, has a single parent, has a dog, loves to hike. Name all those life experiences that give you that rich, meaningful life.

When our personas speak back to the brands, they’re talking through that lived experience. And then across the board, we have a 93 to 95% accuracy in terms of that persona to the actual person.

Can I ask you a very human question? Does your work ever scare you? As you cross the uncanny valley, are you frightened of what’s possible?

This sounds like a canned response, I don’t mean it this way, I am actually really hopeful. Because I feel like as there’s this race with AI sometimes to get to the most efficient, leanest, meanest machine-led sort of solution.

And what I’ve seen us do is turn that around and make it a real asset to the humans at the employer or wherever the company is. To actually just 10x their performance and to get to the solutions that they’ve always dreamt of doing, why they’ve gotten to the job in the first place.

So I actually think it’s a really hopeful turn. So far I’m not too scared about what’s going on.

As you work with clients, do you find that they are reaching insights they would have had before faster, or are they discovering things about their audiences that they didn’t know before?

I think those two things go hand in hand. There is a great notion here between speed and the insights being delivered and whether those two things can coexist.

There’s this 10-80-10 rule where there’s a human at the beginning, there’s an accelerant with AI and there’s a human at the end deciding what to do next.

That process gets pushed down to hours, and now I can do another loop and I can loop it again. As opposed to traditional processes where I might wait months to get the research, go to another phase where we’re collecting for a brief, go to another phase where we’re deciding what to do and then we’re validating at the very end.

Now we get validation throughout the process, which means everyone can dive deeper, faster, independently and come together and really make a conscious decision about, wow, this feels like it’s going to be really disruptive. And now we have the validation to say we can do it.

Obviously, privacy is a big factor as we talk about AI and data. How are you ensuring privacy-consciousness as you build with people?

One-hundred percent. I think the security that we offer in terms of a SOC 2 (System and Organization Controls 2) compliant dataset, when you put your data in, it’s not training our model. It’s not changing the data. [We’re] COPPA [Children’s Online Privacy Protection Rule] compliant, all the different security measures we can put in, we’re going to put in place.

We wouldn’t have clients like JP Morgan and Unilever if we were going to be loose and fast with security or privacy.