The Augmented Human: Building a Philosophy-First AI Strategy with Kerri O’Neill, Ipsos UK & Ireland

John Emmerson: I’m joined today by Kerri O’Neill, Chief People Officer at Ipsos UK and Ireland. Ipsos is a business whose product is human insight and judgement. In the world of AI reshaping how people work and think, how does that impact what you do at Ipsos and your particular role?

Kerri O’Neill: What’s most amazing about Ipsos is that it’s over a 50-year-old business. When you look back at its growth over that period, its biggest jumps come at the same time as the big tech advancements. Ipsos started largely with face-to-face interview processes — stopping you in the street with a clipboard, knocking on the door. Then came the telephone, then the internet, and now AI. So I’m actually excited about what technology and the human interface does for us in terms of our knowledge and understanding about how the world works. That sits at the heart of what Ipsos tries to bring to the world: the voice of what’s going on in lots of different spheres. AI is actually a really exciting technology for us, which is why we adopted it early for all our members of staff.

John Emmerson: Tell us what you’ve been doing. How have you been adopting and embedding AI within Ipsos?

Kerri O’Neill: Often when we talk about adoption, we start with the technology — training everyone to get to grips with a new tool. What’s actually interesting about what Ipsos has done is it started with a philosophy. And this has really helped people to understand, embrace and experiment, because they know the frame by which we want to work with this technology moving forward. The philosophy is very simple to understand and remember. It’s the three T’s: trust, transparency and truth.

Kerri O’Neill: Trust is absolutely core to our brand. So the way in which we use AI cannot break that trust — it cannot be used to shortcut something that then means the data is not valid. Everything around what we do requires statistical validity, robustness, genuine evidence. So almost this philosophy: if this does not create trust, we will not do it, or we will not do it at the speed of its potential. Transparency is about making sure we understand what the technology is doing at all times — how an algorithm works, how we can audit it all the way through. It’s also about honesty in the human layer: what we’re trying to do with AI, what it can do, what it can’t. And truth is really important to us: making sure AI still grounds itself in real experiences and what people actually say. There’s a lot of potential with AI to summarise lots of people’s verbatim comments, and we want to make sure any insights or conclusions made ground back into what people actually said.

Kerri O’Neill: Those three philosophical tenets lay at the heart of how we’ve started to approach AI, and we laid that out before we got to: here are the tools, here’s the prompting piece. Then in terms of training: first of all, we gave AI tools to all staff members, irrespective of role and geography. In part, that was because we recognise this is going to be a really vitally important technology. You wouldn’t not give someone the internet. You wouldn’t not give someone electricity. So we’ve given everyone AI access and trained everyone in the basics: what is an LLM, how do you prompt, how do you prompt properly. We’ve all gone through the phase of doing a slightly expanded search request rather than a proper prompt. And what’s exciting now is that a lot of organisations are helping people create digital twins or second brains — that’s really the next exciting area of AI adoption. How do we go from just using it to offload activities, or using it to be a bit quicker, to actually shaping AI around us? That’s the augmentation strategy. And we have literally just announced, only in the last few weeks, a new strategy around augmentation.

John Emmerson: What I love most about what you were saying is the link back to human involvement and augmentation of human behaviour rather than replacement. Can you talk more about the augmented Ipsos strategy?

Kerri O’Neill: Absolutely — this is in the public domain, we’ve talked to our investors and staff about it. Our strategy over the next few years: we’re calling ourselves Augmented Ipsos. It fits nicely with AI as well. It’s about how we use AI to be an even better version of ourselves for our clients. If I can use AI properly as a working mum, someone who does other things outside of work, and I can use AI to improve that about myself, then I can do those things better. The augmentation element and opportunity of AI is the one I think is genuinely exciting.

Kerri O’Neill: Yes, there are some productivity gains out there. But those are short-term wins. Once you’ve achieved them, if an organisation thinks: okay, I can use AI to cut headcount by a certain amount — well, you can only do that once in your strategy. Augmentation has ongoing potential benefits and ongoing value pools that can be opened up. We have not scratched the surface of the innovation potential in most of our organisations, big and small. I’ve been in HR for over two decades. I’ve seen the cycle where we grow organisations, cut them, grow them, cut them. It’s a really awful hamster wheel to be in as a staff member, as an investor in those companies, as a leader. So I think if the only thing you’re excited about with AI is cutting a little headcount off — your imagination probably isn’t as wide as it could be. There is way more that’s possible here with AI. There are some risks as well, and some genuine things we need to think about, which is why I’m also a fan of a more philosophical approach. If we don’t approach this a little bit philosophically, we’re going to end up making some quite bad short-term decisions.

John Emmerson: AI adoption and implementation within a business has become more of a people and HR challenge than a tech challenge. How does the people function fit into this?

Kerri O’Neill: This is without a shadow of a doubt the most exciting and genuinely significant technology advancement I’ve seen in my career, starting from the late 1990s and early 2000s. When I first entered the world of work, we were in the dot-com bubble phase. One of my first jobs at Norwich Union was essentially: can you train the underwriters and actuaries how to use the internet? Even getting Excel — when Microsoft launched that, it transformed so many jobs. We hardly talk about technology’s role in the world of work, but it’s always been there. So it feels incredibly natural that the people function is involved in technology transformation. I’m a little bit surprised at how surprised people are that HR are involved in this.

Kerri O’Neill: We’re entering quite an exciting time because of the potential for augmentation. I’m talking to founders who say they’ve got a team of ten and none of them are human. Because they’ve asked Claude Code to build the job description and gone off and done it. And some risks and things we need to think quite deeply about. When work can be done literally by anything, we need to think much harder about what we’re going to commit to having our people do. People is often one of the biggest costs of business — or investments, depending on your perspective. We do have to think carefully about making sure we’re extracting all the right value, doing the right things for our businesses. A lot of people have so much more to give. What if AI gives us the opportunity to see more of that? The value of that person in our business doesn’t look so much like getting the admin done anymore. It looks like: what are the relationships they can create? What’s the creativity they bring? So that our business can go faster because we have AI doing this and our human beings doing that.

John Emmerson: In a recent LinkedIn post you talked about cognitive surrender — where people accept what an AI has given them as fact, without questioning or checking whether those outputs are wrong. What do we need to do as business leaders to guard against that?

Kerri O’Neill: I do think this is quite a genuine risk and I’m really pleased to see more academic studies coming out on it. There’s a recent one by Wharton professors called Thinking Fast and Thinking Slow and Thinking with AI. It puts some academic rigour behind what I’ve been seeing and talking about. I think it’s actually quite good that the early models of AI were a little bit ropey, because it made us appropriately sceptical. Of course, technology has been designed by humans and some of this is quite new. So I’m almost pleased there were some ropey answers and we were calling them hallucinations a couple of years ago — because it made us sceptical, and I think that’s incredibly healthy.

Kerri O’Neill: We must always remember that AI is not human. One of my favourite things I’ve been saying over the last two years is: let’s not forget that one of the biggest supercomputers we have sits behind your very own eyes. And that supercomputer compounds when you get people together — it creates connections between those brains that’s even more spectacular. Something that AI is definitely struggling with at the moment. So firms that are really helping people to remember this, to fact-check, to ask AI different questions, to actually click on the sources and check that’s what they connect to — these are some really important behaviours. Not to slow us down hugely, but just to make sure you’re not going to embarrass yourself by putting something out into the world that’s not true or accurate. At Ipsos, this is something that’s really important to us — back to our philosophy around truth and trust. Cognitive surrender, this idea that you just believe everything AI says, is a bit like not knowing you should be careful when a stranger calls claiming to be your bank. Firms and organisations have a responsibility to make sure that people understand AI is not a human being, it doesn’t have the same level of thinking happening. Not every organisation is necessarily doing that training, but I think they should.

John Emmerson: In terms of treating AI well to get the best out of it — some people say treat it like a human, give it context, style, what you want. What’s your take on that?

Kerri O’Neill: I would go a little bit further. First, on the point about early models putting people off: it’s a shame, because AI has improved hugely. You almost can’t recognise today’s models compared to what they were producing even two years ago. A healthy caution, a healthy curiosity about what AI has come up with, a healthy degree of scepticism — not a neuroticism about what it’s producing — is probably the way to go. Just like you’re not going to slam down the phone to everyone who rings just because you think it might be a scammer.

Kerri O’Neill: On the second point, I need us to actually go further than just treating AI like an intern or a graduate. We forget as human beings how much our cultures and environments and surroundings give us through things that aren’t even spoken to us. We are our culture, our environment, our surroundings, the people we interact with — even at different relationship levels — they’re all giving us huge amounts of information that we can contextualise and use to operate. Statistics are quite well out there that we take on something like two billion data points in any given day. So actually I always say you have to go even further with AI: not just talk to it like a grad, you have to tell it more about the context. The people who are creating second brain or digital twin type training are finding it’s actually quite surprising how much you have to say — you have to tell AI your hopes, fears, dreams, visions, your thoughts on tone, communication, timing, style. What we’re really doing there is filling in the gaps for AI around things that as humans come to us through more cultural and environmental signals. That’s a different skill set to what most of us have had. So all of us, no matter where we are in terms of stage and age, have got to go on a learning journey with AI. We’ve got to keep sharing what we’re learning, keep understanding what this looks like, in order to be really competent at using this technology to its fullest.

John Emmerson: One final question: you’re a CEO or leader in a mid-market business watching this back. You don’t have an AI strategy and you have a team that’s a mix of enthusiastic and resistant. What’s the one piece of advice and the one thing you’d advise them to do first?

Kerri O’Neill: Don’t just say to people: go play and experiment with AI. I think it’s really important to go on a learning journey together and to do some of the learning together. Don’t just say to people: go and do that course on their own. Let people maybe do a course on their own with headphones in, but then create dialogue around what you’re learning together. The more we can be social and peer-learn, the better. I’m a big fan of what Trinny Woodall did — she took her whole company, which is a fast-growing start-up with lots on social media, and took everyone out for two days of AI bootcamp to learn how to use it together. They upskilled together, they learned together, no one was left behind.

Kerri O’Neill: We’ve done something similar at Ipsos, not quite company-wide because we’re many multiples bigger, but we have created team-based experiences around learning about AI. What that helps you do is make sure you don’t just end up with a few fast movers over here doing the AI-type jobs and some laggards. We need substantial upgrades across all of our different firms. This is the one time where it’s worth maybe doing something out of the norm — maybe half a day out together to learn and upskill. That could be one of the best uses of your time. If anyone goes and does that after hearing this, please come and tell me what you found. I’m almost certain you will guarantee a shift that you wouldn’t have got if you just let everyone learn and do it in a quite individualised way.

What are the three T’s of Ipsos’s AI philosophy?

Ipsos built its approach to AI around three philosophical principles before introducing any specific tools or training. Trust means AI cannot be used in ways that compromise data validity or statistical rigour — the core of what Ipsos sells. Transparency means the organisation understands what its AI systems are doing at all times, can audit them, and is honest about what AI can and cannot do. Truth means AI outputs must always ground back into what people actually said and experienced, not drift into flattering generalisations. Kerri O’Neill is clear that these principles came first, and the tools and training followed from them.

What is Augmented Ipsos and why did the company choose augmentation over efficiency?

Augmented Ipsos is the company’s recently launched AI strategy, built around the idea of using AI to make Ipsos a better version of itself for clients — not to reduce costs through headcount cuts. Kerri O’Neill argues that efficiency-driven headcount cuts are a one-time lever: you can only do it once in your strategy. Augmentation, by contrast, has ongoing potential benefits that compound over time. She is direct that leaders whose primary excitement about AI is cutting a little headcount are thinking too small. The innovation potential in most organisations, she argues, has barely been touched.

What is cognitive surrender and why is it a risk for mid-market organisations?

Cognitive surrender is Kerri O’Neill’s term for the tendency to accept AI outputs as fact without questioning, checking or applying judgement. A Wharton study she cites — Thinking Fast and Thinking Slow and Thinking with AI — provides academic backing for what she had been observing in her own teams. The risk is particularly acute because AI outputs often look authoritative and are phrased confidently even when they are wrong. Her view is that organisations have a responsibility to train staff explicitly in AI’s limitations, to build habits of fact-checking and source verification, and to ensure people understand that AI is not a human being and does not think the way a human brain does.

Why is treating AI like an intern not enough?

Kerri O’Neill agrees with the advice to treat AI like a new graduate — giving it context, instructions and style guidance — but argues this does not go far enough. The reason is that humans absorb vast amounts of contextual information from their environment, culture, relationships and surroundings without being explicitly told anything. AI does not have this. Everything it does not know, you have to actively supply. People who are building what she calls digital twins or second brains are finding just how much that requires: communicating hopes, fears, communication style, tone preferences, timing, values, and the particular priorities of the organisation. This is a new and learnable skill, but one that most people are still at the beginning of developing.

How should mid-market leaders approach AI training across their organisations?

Kerri O’Neill’s strong advice is to learn together rather than individually. Rather than sending people off to complete an online course alone, she advocates for team-based learning experiences — structured time where groups of colleagues learn, experiment and discuss AI together. She cites the example of Trinny Woodall, who took her entire company out for a two-day AI bootcamp. At Ipsos, team-based AI learning sessions have been introduced across the business. Her argument is that individual learning does not produce the cultural shift you need. The social, peer-learning element changes behaviour in a way that solo courses do not. She is confident that any business that does this will see a noticeable shift that they would not have achieved through individualised learning alone.

How does Ipsos balance AI adoption with protecting data integrity?

Data validity and statistical robustness are fundamental to Ipsos’s business. Any use of AI that shortcuts or distorts that validity is, under the company’s three T framework, simply not permitted regardless of the speed benefit. This includes ensuring that any AI-generated summaries or insights are verifiable against the original verbatim responses from research participants. Kerri O’Neill frames this as a commercial as well as an ethical requirement: Ipsos’s clients pay for the integrity of its research. AI accelerates the need to have strong governance around data, but does not change the underlying standard that data must meet.

Why has the HR and people function become central to AI adoption?

Kerri O’Neill argues that technology transformation has always required the people function, even if HR’s involvement has not always been visible. She draws a direct parallel to earlier technology shifts — the internet, Microsoft Excel — where changing how people work was as much a cultural and behavioural challenge as a technical one. With AI, that dynamic is more pronounced than ever because the technology directly affects cognitive work rather than physical or administrative tasks. The questions of what people should be doing, how to bring them on the journey, how to manage anxiety about job security, and how to build new skill sets are all fundamentally people questions. She is direct that HR involvement in AI adoption is not a surprise — it would only be surprising if HR were not involved.