Adrian Blair, CEO & Co-Founder, TrustPilot: Winning Customers. Trust in the Age of AI

So, just to introduce myself, I'm Adrian Blair. I've been the CEO of Trustpilot for the last couple of years. I'm going to be talking about trust, winning customers trust in the age of AI. And the reason we care so much about this at Trustpilot is that we operate the world's largest independent customer feedback platform. So we have millions and millions over 350 million human experiences with businesses that people describe on Trustpilot. And we think this is enormously valuable for building trust in the age of AI. So let's talk first about trust itself and what it actually is. And let's think about the development of AI and how that's going to affect it. We think of trust as a confident relationship with the unknown. So trust is what gets you to do something that you're not absolutely certain of. Like you know 2 plus 2 equals 4 does not require trust but buying a ticket from an airline and getting on that plane and crossing the Atlantic requires you to trust quite a lot of things. So trust is a confident relationship with the unknown. That's how we like to think about it. And that came from an author called Rachel Botsman who wrote extensively about trust.

But AI is evolving and changing how people actually build trust. If you think about the history of this, it started out trust is a very local thing. It was who could you, who did you know who was trusted in your community to be an expert on something or to be particularly reliable. It evolved particularly in the 20th century into something that resided in institutions. So you had big brands with big budgets and that's how people sort of decided what they could trust at a global institutional level. Then with the internet and the development of social media, it became much more distributed. So everybody could speak to each other and you could assess trust by listening to the voices that were out there and try and figure it out that way. We think with AI trust is going to become embedded. Trust is going to become embedded in AI systems. So we're moving into a new era for how trust actually works across society.

AI is moving trust from not just something that exists in the background, but something that's actually driving decisions that AI technology is going to be making. And what we're determined to do at Trustpilot is to make sure that that's rooted in genuine human experiences and what's actually going on out there in the real world. So of course what's going on right now is AI is developing from a kind of chat interface where you ask it questions where you have a conversation to something that is out there in the world taking action. So AI commerce is becoming a thing. AI agents are out there in the world taking actions on your behalf, doing what they're instructed to do. But if we're not careful, they're going to do that without actually taking trust into account, without actually using trust signals. And we think that's potentially quite a dangerous position for these AI agents to be taking us.

This is an example of where previously AI might have told you who the top banks were or you might have gone to Trustpilot to find out which banks had the highest ratings but AI agents are going to go a step beyond that and say right we've switched your account you're actually now banking with someone else because you asked me to take that action on your behalf or it could be buying something or it could be changing who your utility provider is. So AI is just becoming increasingly powerful. We think the effect of that is that trust becomes even more important. So the more powerful AI gets, the more important it becomes that trust is actually embedded in these systems. The problem is right now people don't trust AI to get it right. And that's quite understandable. I actually asked one of the leading frontier AI tools recently how often I could trust what it said, and it told me the answer was 37% of the time. Which didn't exactly inspire trust. So we have a bit of a crisis of trust right now as it stands in AI. When you ask people in America, do they trust it? 54% of them tell you that they won't trust it to actually take decisions on their behalf. So I think we're entering the situation where technology is able to become agentic and to go out and take actions on people's behalf. Technology is becoming capable of doing that. But do people actually want it to do that? Will people actually trust it to do that on their behalf?

What are the problems people associate with it? Well, it's difficult often to know the source, right? So AI might do something. It might take some action, but it's not clear how it's actually decided to do it. It might look very confident, like when the AI told me that it was right 37% of the time, but how do I actually trust that figure? How do I know how do I know what I can actually trust? How do I know ultimately whose side it is on? Is it there for me or is it there to serve its creator somehow? So we have multiple issues around trust in AI in the world right now. So we think what's crucial is that AI systems interact with humans and interact with real human experiences. So the scale of human experience is absolutely vast. People are having billions and billions of experiences every day with businesses across the planet. We think AI needs to somehow ingest that and learn about those experiences and about what's going on. And that's where we think at Trustpilot, we've got an important role to play. Every day we get more than 200,000 bits of feedback left on Trustpilot. So those are people describing their experiences with businesses over 350 million historically and obviously that number is growing every single day. So it's growing organically. We're getting far more feedback over time.

What we're now launching is a way to embed that and understand that feedback in your AI product. So we're launching something called Trust Layer which is essentially an API that gives access to all of that feedback. So that vast repository more than 350 million voices that have been left real human experiences described on Trustpilot. We're launching an API that gives folks access to that so that they can embed it in AI systems and have agentic AI that's actually listening before it takes action, listening before it speaks. And obviously the more feedback that we get on Trustpilot, the more businesses we're working with, the more effectively we can do that over time.

There's this fantastic quote from Sam Altman who whatever you think of him he knows a bit about how to grow businesses and he said the best companies in the world grow by word of mouth having users who love the product so much that they tell their friends and that's essentially what's going on on Trustpilot every day. We get generally it surprises some people particularly here in France but we actually get more positive feedback than negative feedback on Trustpilot. So there's a lot of people recommending things to their friends on Trustpilot. And we think that the more that continues, the more people are using this open platform to recommend things to each other, the better job we can do at embedding this in AI systems through this trust layer product that we're launching and making the future of AI something that is actually built on trust and therefore built to last. So overall, we're excited by the direction the world is moving in. You know, we think with AI agents, they have huge potential, but only if people are actually going to use them. And in order for people to use them, they need to be trusted. And in order to be trusted, they need to be learning from real human experiences, not just inferring whatever they can from data, but actually real human experiences with businesses need to be at the center of it. And that's all from me. If there are any questions in the room, I'm very happy to do what I can with them. Yes, sir.

Well, I think the key thing is to make sure it's a great question about how do we ensure that the agents are actually handling this in a trustworthy way. So we have a great responsibility to make sure that the content on our platform is really trusted content. So we actually have one of my direct reports is our chief trust officer. So he runs the team responsible for making sure that the content is trusted. We use technology. We have our own AI systems that look at every review that comes into Trustpilot. There are hundreds of metadata points that describe it. So we look at all of that data to decide what's actually trustworthy. And last year 2024, we took 4 and a half million reviews off the platform. So every day as those 200,000 reviews are coming in, we're taking lots and lots off as our AI systems tell us whether it's trustworthy or not. We also have a flagging system, so people can flag content they don't think is reliable, which we then go and investigate. And we go after bad actors proactively. So we have our own fraud investigation enforcement team and we're going out for example to take action against review sellers who try to make money out of scamming systems like ours and we always want to be on the front foot in taking action against them. So yeah, we're doing a lot on that front. You had a question too.

So on Amazon India, you'll see that many of the products there are very few good reviews and then mostly 4, three and two star reviews are like the highest percentage and that tends to be the case with restaurants as well. So how can AI, how do you see AI accounting for this cultural bias where certain markets have a tendency perhaps towards more critical reviews or more negative reviews? And I've had this happen in my own business.

So well I think this is one of the great things with Trust Layer. By making this information available through an API, AI systems will be able to look for themselves at what the review scores are coming out of each geography and be intelligent about exactly as you say correcting for that and realizing that if a review in one country doesn't mean the same thing as a review in another country because of the cultural nuances then it should draw its own conclusions from that and treat them differently. But I think it's a great thing about putting this raw information in the hands of the AI companies, that sort of gives them the opportunity to really listen to what people are saying and listen to real human experiences and all the nuance in that across geographies.

It's a great question. Any more questions? Yeah.

Yeah, I mean you're absolutely right. It's already much better than what you would get by just scraping. It contains more than you would just see on the public web, but obviously we're looking at ways that we can augment it over time and make it better and better for people. But look, at Trustpilot, we want people's voices to be heard. So all these reviews that we're getting, we love making them visible and available out in the world in different places. So we're not fundamentally an advertising business. So we're not all about how many people look at it on our own website. We actually love it if reviews are having impact elsewhere because we want people's voices to be heard and we want the feedback to be listened to. We also want the businesses who have great feedback, we want them to benefit, right? We want to help them look great in lots of different places.

Any more questions?

Well, look, I think the big thing is Trust Layer. I mean that's our big AI launch for this year. We launched a lot of other new products for businesses earlier this year. We just launched AI review summaries. So if you look on a lot of our Trustpilot profile pages now you see the AI summary of the reviews of that business. If it's a business with like a million different reviews that's a very important bit of information. But the main thing is Trust Layer itself and we think there are many different categories of business that have potential to benefit from Trust Layer. So it could be a large language model. It could be an investor who's looking to do due diligence and wants to understand customer feedback on companies for that purpose. It could be a consulting firm who wants to understand what's going on within a sector. Maybe you want to know how you stack up versus all your competitors and you can see that through the system. So there's all kinds of different use cases for us and that's why it's such an important launch for us. All right, thanks everyone. Cheers.