The idea that human research is replaceable by machines is simply wrong; the machines really need the humans to supply that data.

When Stephan Shakespeare and Nadhim Zahawi formed YouGov in the United Kingdom in 2000, data was the beating heart of the operation. Today, with a global presence across Australia, France, Germany, Italy, Singapore, Spain, Switzerland, and the United States, as well as the UK, that remains the case.

Based on the claims of the AI hype cycle, it might be thought that YouGov was precisely the sort of organization ripe for usurpation by the rise of the tech. The firm specializes in market research and opinion polling through online methods, sourcing data from an invited group of Internet users, and then weighting these responses in line with demographic information. That’s something any aspiring bot should take in its stride, no?

But according to Shakespeare, the threat has been greatly over-rated:

We've got a two percent increase in our data products. A while back, people were thinking that's all going to be taken over by AI. Actually, AI is a big benefit to us and a big opportunity. It isn't something that's been holding us back...There are particular anxieties around our industry because of AI and will we all be replaced?.

Humans are not about to disappear, he insists:

I actually don't think they will even necessarily reduce. I think that we will grow in scale and be able to do much more with the people that are there because the handling, the helping of clients around these flows, getting the most out of the system is something that remains very important. I just say that to ease concerns that people might have at the company that the jobs are going.

YouGov’s ‘special sauce’ is human in the form of its people panels, argues Shakespeare:

We have very strong open relationships, transparent relationships with our panelists. They trust us in a way that they might not trust a pure AI company that does other stuff with their responses. We are very clear about how people's data is used. We give them lots of assurances that we can then follow up with very ethical behavior. I'm not saying others aren’t ethical. I'm saying that they're not able to develop those relationships. Our ‘special sauce’, our moat, is that relationship with large numbers of humans sharing a lot, not just doing a few surveys, but actually sharing data, sharing even downloaded behavioral data.

Our moat is our panel. Nobody can touch that. Nobody can equal it. The entire desire of the synthetic companies is to squeeze out humans. They go and they buy mass polling and hope that they get this. They make their [digital] twins and they don't have ask to talk to them again for another year and get lots of value out of it. That doesn't work. They have to have live data to make a great product.

Overall, he says

It is the opposite of what people expected - AI is very much our future.

And YouGov is embracing that future. The firm is looking to AI to enhance its service proposition, citing an agent that can tell you about what's in the data and help you find more. Expect more of that sort of thing to come, he says:

Everybody has that from an agent that looks at your data. We have a heck of a lot more data than anybody else by an order of magnitude. And to get the most out of that is what we're doing, we’re adding MCPs, but above all, we're making it a platform that an agent can use directly. Those are our AI-powered solutions and the platform we've long talked about, and now it is becoming really a full-on data platform.

The sheer quantity of data at YouGov’s disposal is a critical differentiating factor here, posits Shakespeare:

Where it becomes exciting, is that the agent can not only say here's what the data says, but do you need some other data, we can get it for you within 30 minutes. And you can write your own questions or you can have the LLM (Large Language Model), your own LLM or ours or whatever we use in that place to write the surveys for you from all the knowledge that we have to answer your questions.

And there’s a lot more ambition here to be realized, according to Shakespeare, as the firn looking to make an end-to-end automated research flow a reality:

The advantages are higher quality in some respect in terms of speed, consistency, accuracy, all of those things, because humans make mistakes. The more of that process that we can automate, the more of it we can streamline before it even gets into the system, [the better].  At the moment, it's typical that people come with their projects and they design them in any way that they want, and then we make that happen and it's highly customized. But the advantages to them as well as to our own efficiency is to streamline how they think about the work and how that works its way through the system and how that helps to reduce error.

Not that YouGov has aspirations to be “advanced creators of AI”, says Shakespeare:

We're not saying we're going to be a gigantic AI company. That's not at all what we're saying. We're talking about in our industry, focused on how clients of market research use data in their businesses, that’s our expertise. That's where we're focused on, and that's where we want to create the most advanced agentic model that there is, which we can do because of all the things that we already have here. This model of being able to look at the data we have and instantly verify and add to it is from the same people is, I think, unique in the industry, certainly at the scale that we can do that. 

But there’s also a pragmatic realism about the limitations AI has in the market in which YouGov plays. Shakespeare explains:

Most of what we do is to measure change. We obviously do some static polls and studies and so on, but that isn't the core of what we do. We measure change day in, day out, by lots of different products and by big custom trackers. All of those things are about understanding how today is different from yesterday - and the one thing that AI and synthetics cannot ever do is tell you how things are changing, except retrospectively.

I think a lot of synthetic data companies like to tell us they can [project forward] because they're so brilliant and they know how the weather is going to affect everything and all these other mysterious things. If they could do that, they could only tell anything about the future by continuation of patterns from the past. They might say, ‘If things continue as they look today, it will go this way’. That's not why you buy trackers; you buy trackers to see what actually happens.

As for other “advanced creators of AI”, YouGov has allied with two firms that might have been perceived not long ago as competitive threats - Anthropic and OpenAI. Shakespeare says:

They’ve become big clients of ours. We've done £4 million just in traditional reputational work that they use to understand audiences, but also to communicate. Anthropic has made a big deal out of very large surveys that we've done for them, which connect them to their users and to the public in one of the ways that is central to their strategy. Their spending there is already the biggest growing part of a specific bit of our business.

There is  a degree of client expectations management about what AI will be able to achieve that needs to come into play here, cautions Shakespeare:

Clients are expecting to get more for less. I mean the basic assumption with AI is that we are already benefiting from the magic of it. Obviously, we're not. We are not at the point yet where we have the savings that come in. I think this is true around the whole world in all industries. It's for most suppliers, they're not yet getting the advantages of the efficiencies which the clients are wanting to bake in.

He adds:

One of the funny things about it is that those clients come back to you and tell you how you should be doing a business. They say, ‘Well, actually, we've looked at how you structure your things’ and say, “Why didn't you do it this way? It would be much faster or much better.’. They can be very demanding in a good way. It forces us to innovate faster and to listen and to see how we want to deal with that.

Our consultancy, I think, is going to be assumed to come free because actually, they don't need it, and they don't often even care how you're packaging up the data. In fact, they tell us. ‘We just want it more’, so yes, this is changing the demand. But for us, this is what we want. Where we make our margin is in quantity of high-quality data. The bit that they can't get from anyone else is the actual relationship with humans. That is a trusted personal relationship actually and you need that, I think, in order to create great data.

An interesting and pragmatic assessment of the pros and the cons of AI on an established business model. How YouGov is adapting and evolving to the rise of the tech is a good exemplar of how to embrace the future without a wholesale dismissal of the past.