How VCCP made an anti-scammer AI granny for O2

VCCP Faith’s Ben Hopkins and Morten Legarth on using AI to boost human creativity

Daisy is O2’s new head of scammer relations. She was created by VCCP’s AI agency, Faith, to take calls from suspected fraudsters, keeping them on the line for as long as possible to protect real customers.

When we last spoke to Faith’s joint creative director Ben Hopkins in March, he teased the agency’s ambitions to blend human ingenuity with AI precision. In this interview, Hopkins and his creative partner Morten Legarth, discuss how Daisy was brought to life, the viral impact of the campaign, and what her success tells us about AI’s evolving role in advertising.

Why an AI granny, and what made her the perfect fit for O2?

Ben Hopkins: Faith was set up to find positive uses of AI at a time when there was still a lot of trepidation surrounding the technology. Morten is a big fan of scam-baiters on YouTube, and he had this idea that AI could supercharge what they were trying to do.

Morten Legarth: So why a Granny? Scammers tend to consider elderly people as vulnerable and easy prey. While many elderly don’t fit that mould, we felt a lovable Granny would be the perfect weapon to teach them a lesson. Not only would scammers suspect Daisy less, the scammers’ perceived notions of the elderly being slower or getting confused actually helped to mitigate some of the limitations of technology, such as latency.

What makes Daisy, Daisy? How does she get in touch with the scammers?

BH: Daisy has various tactics to maximise her ability to waste scammers time. One is number seeding, where we take numbers and spread them around the web by inputting them into lots of data capture forms like online competitions on questionable websites. These put numbers on what are known as mugs lists which are bought and sold to scammers call centres to call. We also have access to numbers sent to 7726 from O2.

ML: Daisy is a combination of different AI tools and services. In this rapidly evolving space, some of the tools that we started with to prove the concept were different from those that power Daisy now, so we could keep improving her performance. In terms of how she works as a character, it’s a combination of knowledge and personality training. For knowledge, we were lucky enough to work with Jim Browning (who is famous in the YouTube scam-baiting community). Daisy is trained on anonymised call data and techniques and scam formats used by phone scammers.

BH: It also became really clear that we didn’t just need a voice for Daisy, she needed a whole world building around her: her personality, her interests, her family. Jim made it clear that tangents were really important to frustrate scammers and keep conversations going. From this, we drew inspiration from our own grandparents. Morten’s gran was an avid knitter, mine made a formidable lemon meringue. Not only did this detail make Daisy more believable, it made her the ultimate filibuster to waste scammers’ time.

How much of Daisy’s appeal tapped into collective frustration with fraudsters, and how did you balance that with humour?

ML: Most people have their own experiences with scammers. Seven in 10 Brits have been on the receiving end of a scam attempt, so we knew this would resonate with a large audience. It’s about getting revenge, which people can relate to, but also we’ve a really important message to deliver on this, to get people to think about what the person on the other end of the line is trying to get them to do and to hang up and report them to 7726 if something feels off. To balance this we leaned into the same type of humour we’ve seen in the YouTube Scam-baiter genre to make the content entertaining while having a clear call to action.

BH: You obviously can’t predict what the other person is going to say or do, so you don’t know if something funny will happen — some of the calls were really horrible. We had seen lots of scam-baiting videos, though, so we did know that those entertaining conversations take place. I don’t think Morten or I will ever forget the first time we let Daisy loose on a real scammer. It was surreal. It worked straight away and the things Daisy came up with were hilarious. Once we’d got the calls and had captured content, it was then a matter of selecting the best clips to give people good insight on how scammers operate, demonstrate how Daisy works, and entertain people enough to watch the whole clip.

Ben, in March you described Faith as ‘AI assisting creativity, not replacing it.’ With Daisy, where did AI lead and where did humans step in? How key were personal and external influences in shaping her design?

BH: Assisting creativity or humanity rather than replacing it is exactly what we’re doing here and it’s what makes Daisy such a good use case.

ML: AI is just a tool, really, like a camera or a paintbrush or a piece of design software — it has a creative use. One of the problems there has been with AI is that some people have put the tool ahead of the use case, so we’ve had some gen AI-led campaigns that are using AI just to say they have — where the tool chosen is the only creative thing about the campaign. Something like this scam-baiters project is only possible because of AI, Daisy can take multiple calls at once, she can run 24/7 if needed and improvises endlessly. We’d struggle to have the same impact using a human actor.

BH: It’s not to say that humanity didn’t play its role though. What makes Daisy so believable is that we used a real voice as the AI tools weren’t convincing enough. We used the real grandmother of a VCCP staff member to record a guide for testing and it just stuck really. She simply was Daisy. Interestingly we did approach professional voice talent as we assumed this would be the path we needed to take, but again, we met with hesitancy and trepidation, which we totally respect. It’s going to take time and the formalising of best practices to ensure that AI and human talent can intersect in ways that preserve and respect the value of human creativity.

It’s clear from the media coverage that Daisy’s made an impression. What’s been the most memorable reaction so far?

BH: I think the first reactions of people who have spoken to Daisy have been incredible — we forget sometimes because we’ve been working on this project for a long time. We‘ve had some journalists speak to Daisy and freak out a bit at how real she sounds.

ML: The press reaction in general has been amazing. We’ve had coverage from all around the world, seeing Cat Deeley and Dermot O’Leary discuss Daisy on ITV’s This Morning was surreal but a great complement to the overarching sentiment that so many real people see this project as being such a positive application of AI.

How do you see public perception of AI evolving through campaigns like this? Is there a shift from AI being a tool of fear to one of fascination, even affection?

ML: Absolutely. I think a lot of people have a perception of AI as really negative. Either that it’s going to steal people’s jobs or even worse. No doubt AI will disrupt the workplace massively in years to come — in the same way that the creation of the internet did. So the fear is understandable and in some cases justified. But there are positive benefits to AI — it can supercharge human ability, giving us extra tools to do things that just weren’t possible before, it’s not just about efficiency or time saving. We’ve had so much feedback about this campaign that it’s refreshing to see a positive, tangible use of AI. I’m sure there’ll be many more to come in the future.

Can we get a sneak peek at Faith’s next big project —and might we see another quirky protagonist like Daisy?

BH: I’m afraid you’ll have to wait and see! We’ve had a lot of interest off the back of this project from people who are seeing the potential of AI, but we’re not necessarily looking to use the same solution for every problem. Building effective communications is more about assessing the problem and then designing a creative solution, so the answer won’t always be Daisy.

I think it’s really rewarding when we have a client like Virgin Media O2, who have worked with us on the development of more than one AI solution now, and they share our approach to AI and place faith in us to develop something new. Our VCCP colleagues at Bernadette have also built a great tool, My Cadbury Era, recently. It shows that there are some really savvy marketers out there who are embracing the potential of gen AI and getting good results.

Natasha Randhawa, Editor-at-large at MediaCat Magazine

Tash joined the magazine as Editor-at-large at the start of 2023. Previously she headed comms for The Marketing Society (2018-2022). Now, as Editor-at-large, she travels around Southeast Asia, writing about culture, social impact, creativity and technology, and how these forces influence the marketing industry and wider business world.

All articles