Ben Telfer:

Hello, everybody, and welcome to today’s ICMIF webinar, Practical AI Applications for a Modern Mutual. Delighted to have so many of you joining us today. AI is obviously a topic that we’re seeing a lot of interest in from our members around the world. So we’re delighted to be joined today by Steve Ferrari, who is Chief Transformation Officer at OneFamily in the UK. Steve is going to be sharing some practical ways that OneFamily is using AI. And we’re delighted he’s going to share some use cases that are already created customer value for OneFamily and its members. Steve, welcome. Delighted for you to join us today and I’m pleased to hand over to you.

Steve Ferrari:

Hi, Ben, and thank you. Yeah, so my name’s Steve Ferrari. As Ben said, I’m the Chief Transformation Officer at OneFamily. OneFamily is one of the larger financial mutuals in the UK. And I’d like to spend the next 40 minutes or so just telling you a little bit about OneFamily’s experience with AI and how we’ve delivered customer experience and efficiency savings. And over the next 40 minutes, what I’d like to do is give you a brief introduction to OneFamily and myself and then why AI? Why AI for OneFamily. A little bit about we started, how we set it up and what we’ve done, and then really importantly, what are the benefits, how we deliver the benefits from the introduction of AI. And then finally, just some key takeaways, hopefully some pointers for you how you might want to implement AI within your organisations.

So, OneFamily, who are we? We were established back in 1975. So we’re almost 50 years old. I like to think we’re a trusted provider of financial products for children and adults. Quite important on the children side of things. I’ll say a little bit more about that in a minute. We’ve got 2 million customers in the UK and we look after about five and a half, just over 5 billion of funds under management. And what’s really important to OneFamily and especially to myself, is the fact that we really pride ourselves with the service that we provide to our customers and members. And our core service goals are always in the 90s. So really good focus on customer experience. And that’s really important because when we talk about AI, certainly for me, for OneFamily, the key thing is we cannot let customer experience suffer through these implementations. What we need to do is implement them well, where we see the customer experience actually improve and we deliver efficiency benefits as well.

In terms of our business, over half of our business relates to a product called Child Trust Funds, which is a savings product, where many years ago the government actually provided some government money for parents for the children’s savings. That’s important. That’s over 55% of our policies. And that’s where we focused most of our AI efforts. So we really went initially to where we would get the biggest benefits from any AI investment. And then about 23% is other savings products. In the UK these are called individual savings accounts and a lot of them are for juniors for the under-18s as well as our junior bonds. So that’s the focus on child savings. And we also have a protection business as well as a small lifetime mortgage business as well. So that’s been about OneFamily. Myself, I’ve been working in financial services for over 30 years, 11 years here at OneFamily and previous to that with companies such as AXA or Prudential and PricewaterhouseCoopers.

So why AI? I like that first saying at the top. A few years ago, back in 2020, it was very much a case of, I had a problem and I was looking for a solution, rather than, I had a technology that was searching for a problem. Which is sometimes what you get when you get new technology and it’s driven by IT. And actually looking for ways that they can use this technology. Actually it was the other way around in OneFamily, you see that those four things, pretty classic problems where I wanted to see if I can improve the customer experience, increase revenue, deliver higher productivity, and lower risk. These are my classic problem statements, which I’m sure are very common across all your organisations. So how can I improve on all of those? Back in 2020, searching around, AI seemed to be a really good way of addressing these problems.

So it is very much starting from the problem, what solutions out there to sort that out. So how did we get started? And really it was, what were the use cases that we could use AI for? Where could I get the biggest benefits? So with the customer service operation and tele sales, when you look at the biggest drivers for the cost, which drive a lot of the manual effort for us, again, no surprises there. Having to handle a large volume of incoming calls from our customers and also dealing with a lot of post from our customers in terms of customer letters, application forms, exit forms and things like that. So really some of the key questions that I ask myself is, why are customers calling us? Why have I got such a large volume of calls coming in? And when I say large volume, for us OneFamily, we’re talking about three to 400,000 calls a year.

And really what I wanted was an ability to get a much richer and better understanding of why customers are calling us. You think that’s quite an easy question to ask, but actually if you’re relying on anecdotal evidence or reporting from the agents who are quite often very busy, you’re not really able to get to a really granular level of detail, which is what you need if you’re trying to look at root causes for why customers are calling us up, how to reduce failure, demand and so on. So really I wanted to be able to get some really rich valuable insight from the calls that were coming in. And in terms of the paper in this day and age, why is there so much paper? What could we do to streamline processes, look at digitising them, look at how technology AI could reduce the requirement for customers to write in to us.

So they were the kind of problem statements. And in terms of the outcomes that I wanted to deliver. Again, probably no surprises there. I think everybody on this call would want to see these things. Better customer journeys and a better customer experience, more choice for the customers in contacting us, not just phone paper, but also digital channels, emails, SMS, chat bots and so on. Giving the customers the ability to transact with us 24/7, not just when the call centre was open. And for me, fewer calls, if we get things right first time or things are clearer or there’s more simple, quick automated ways of dealing with certain simple transactions, then that should drive fewer calls, which is good for efficiency. But also there’s lots of customers who would like to transact quickly and probably don’t want to call us up. And that’s particularly relevant for our Child Trust Funds because a lot of the maturities arise when the child reaches 18 years old.

I don’t know if you’re like me, but I’ve got teenage kids and they really don’t like talking to people on the phone. They far prefer to use text messaging services and so on. And also deflecting cause. Could we deflect cause away from humans to automated processes for those simple transactions? But it’s really important there. Certainly our default option is always if a customer needs to talk to us, we always make it easy for them to do so. We don’t hide phone numbers, we make it very visible. It’s only if they want to use the automated channel they can. But if they want to talk to a human, absolutely we never stop that from happening. And higher quality calls. Certainly when you look at measures for customer satisfaction, one of the things that comes up very high is customers being able to talk to knowledgeable call centre agents. So it’s really important that our call centre agents are equipped with the information at their fingertips to be able to give the right information all the time.

Fast turnaround times. Of course, if you digitise things, you get rid of snail mail, then you improve the turnaround times, which is great for customers, great for efficiency. And if you do digitise things, streamline them, use AI, then naturally that will mean less errors if there’s less manual processing, less paper if you’re digitising less waste. So that’s a summary of what we wanted to achieve from looking at ways to use AI to address these issues.

So when I talk about AI. I mean AI has been around for many decades. It’s just recently there’s been a hell of a lot of airplay regarding artificial intelligence and a lot of that relates to gen AI, things like chat GPT. We haven’t used that just yet to any great extent. So the AI I’m talking about is the more traditional artificial intelligence around natural language processing that’s taking natural language or text and creating it into data that the machines can use and also machine learnings. So that’s a software which continually learns through customer behaviour to improve the way it deals with queries with transactions. And so the three, I’d say classic bits of AI technology that a lot of customer operations use, are, one, speech analytics, two, chat bots, and, three, knowledge management. So they were the three technologies that we really looked at seriously three years ago and that we’ve implemented over that period.

So how did we set it up? And I think the first thing to mention is that it was business led. So as I said before, I was looking for a solution to my problems rather than technology looking for a problem. So our programme was very much business-led. It was led by me, I know I’m customer transformation officer now, but that’s very recent. And before that I was the customer operations director responsible for the customer service operation as well as tele-sales. So they’re the areas which are very much customer-facing. So I sponsored the programme and the key person who was responsible for the actual the day-to-day for leading the project was my customer experience manager. And I think that worked extremely well because that person was very well steeped in looking at ways to improve customer experience and had a very good appreciation of the kind of issues that our customers face, very good understanding of the customer journeys and the things that we could do to improve those.

And that person also had a small number of people in what I call a low code team. So low code for me means people who are able to use these technologies who don’t have to be skilled or seasoned computer programmers. These are people that we handpicked from our customer service operations who had a good aptitude for this kind of technology, who are quite analytical, very customer-focused, and had the capability to receive training, which would only last weeks or months, not years, to get up to speed with the technology such as that were related to chat bots and speech analytics and so on. So that was the nucleus of the team. So we started our journeys three years ago.

But where to start? So we thought we needed some external help. So we engaged a company called Alpha, had a lot of experience of doing this with other clients and they were really, really useful in helping us to clarify our thinking on the use cases that would deliver the best benefits and also the type of technologies and the vendors that we could go to. We also spent quite a bit of time with our existing third party vendors to see what ideas they had. So we had a managed service provider that looked after all our telephony operations, the back-end telephony infrastructure delivered by Mitel and our core recording company called Red Box where all our core recordings were stored. And again, they had some very good ideas of who we should be using. So a combination of those two. We scoured the market, we did some market testing, we talked to a whole range of third parties, and on the right hand side of this slide, you see the vendors that we went with.

We did have the option of going to companies that provided a bit of a one-stop shop that provided all those technologies. But as you find quite often that might not be the best solution. Those type of companies may be very good in one or two, but not in all of them. So we decided to go for what we considered best of breed for each of those technologies.

So we went for a company called CallMiner for speech analytics and Gartner, a global company that we use for looking at who are the best technology providers out there. Call Miner was what they call in their magic quadrant. So very experienced, very good at delivering these services. A company called ServisBOT for chat bots. They’re quite big in the States and Ireland and they were moving into the UK. So I had a very good offering there. And in a company called KMS Lighthouse for knowledge management, again, another very well-known company. So they’re the three for the classic AI. Probably worth also mentioning that we also use some other companies for robotic process automation, a company called BluePrism and Microsoft PowerApps have got a very good platform as well. So it was quite a lengthy process to start with. Three years ago, well I want to say lengthy process, that was probably a few months there, well probably about six months to get through that. To get to a position where we knew what we wanted to do, we knew who we’d wanted to go with and we’d selected the providers.

So having done that, then we had some particular use cases that we wanted to use the technology for. So what I’ll do now is just spend the next 10 minutes or so go to go through probably four examples of how we’ve used that technology and the benefits that we’ve actually derived from it. So the first one is using speech analytics. So speech analytics on its own is a fantastic technology and it’s one where it’s invaluable for being able to really get to the bottom of particular issues. It’s fantastic for root cause analysis, it gives you incredibly rich information. And what we’re able to do is to record all the calls that come into our call centre. And those calls can all be categorised and then they’re all transcribed to a very high degree accuracy. So from voice to text, the accuracy is about 95%, which is more than enough for the software to do the analysis. So it’s incredibly accurate.

So how we’ve used this particular example was where in our CTF product, our Child Trust Fund, if you remember it’s our biggest product. We looked at all the calls that were taking quite a long time and we picked out those which were taking more than 10 minutes or so, which for us is probably almost double our usual average handle time in a call centre. So I wanted to find out, well give me all the calls which are of that timing or more, and let’s look at, well, what’s happening in these calls? Why are they taking so long? What are they all about? So we identified that for these ones, because it was a new process in terms of the maturity process for Child Trust Funds, some customers were a sizable number of customers were finding it difficult to navigate the process. They had lots of questions and they needed quite a lot of holding for the digital process.

And our call centre agents naturally very, very helpful and would spend the time taking them through its step by step. And when we looked at that, we found that, when we analysed the calls, the type of issues or queries that the customers had, we could categorise them quite well. And there were a relatively small number of those. So we designed a document which answered all those questions in a very clear and visual way and we were able to, on the call, if customers in future had those type of queries, we could take their email address and send them the document. And that had a number of benefits. One, all the information was there. Secondly, the information could be referred to at any time. And thirdly, quite often the parent with the child who just turned 18 years old for this maturity, they could use the information pack together to go through the process.

So what we found after we implemented that change, briefed call centre agents on this new process, sending that documentation out, we saw that over time after about three months when we measured the impact of this change, we saw that the average handle time had reduced close to 5%, 4.4%. And we could see directly what FT saving was associated with that and that delivered a good efficiency saving. But more importantly, what did it do to customer satisfaction. So we have a call survey after every single call and we could see after three months after launch, the actual satisfaction rating actually increased by from 93.6 to 93.8. So it was really pleasing to see that we made the change, it speeded up the process for us for the customer. It generated an efficiency saving in terms of shorter calls and it improved the customer experience. So for me that was a great example of a win-win.

Another example using speech analytics, we scan the calls to see if there are any common themes where there is any sort of comments around dissatisfaction for whatever reason. And we saw this particular transaction type where for a particular claim we were asking for certain documentation which gave rise to a number of customers asking, “Well, why do you need this information?” And, “Okay, if you need it, what’s the process?” So that generated a little bit of dissatisfaction and extra call traffic. But when we actually dug into it, we analysed the calls and then we went back to the actual process. We found that actually the reason why we were asking for some of these documents was historic. And actually now those original reasons weren’t really valid. So we are asking for things when actually we didn’t need them for the vast majority of the customers. We did in a small number of cases. So we made the change to the process and you can see there on that graph in fairly short order, the volume of calls associated with that dropped off quite dramatically.

And an allied to that, we saw that the complaints associated with that particular issue, they pretty much almost went to zero. So another great example of how we used speech analytics to do the root cause analysis really forensically see what the problem was, go back to the process, made the process change, and then you could see the results in fairly short, in the matter of months of how it reduced the core volumes, increased efficiency and reduced complaints, which no doubt increased the customer experience. If you look at chat bots. So we started our journey in a fairly basic way and it was very much test and learn. So you’re never going to get it right from the beginning. And also you want to minimise any impacts in the early days when you are learning. So what we did is we didn’t advertise the chat bot on many of the pages on our website.

He was only on the contact us page and we used the speech analytics to analyse well, which are the main query types coming in from our customers so we could get a good view of the type of questions that might come in via the chat bot and bought answers to provide. So we tested it. When I say tested it, we would go live on that contact us page for hour or two, then take it down, see what the results were, modify the chat bot and the answers that it provided. And we did that over the course of a number of weeks until we were happy that we were answering a high proportion of the queries that were coming through. So that’s a classic example of test and learn, do it small and then when you’re happy with the results, scale up. The other thing is, when you’re looking at chat bots, ideally what you need to have is also the ability to have live chat as well.

So if someone’s on the chat bot sort of digital, and if at the end of it for whatever reason, the chat bot cannot provide the answer to the question that you want to have the ability of transferring that query via live chat to a live agent. So it still remains within a digital channel, but you are providing that escalation route to a live person. So that’s really important when you are thinking about chat bots because if it’s just chat bot on its own and you get to the end of your query and it cannot be fulfilled and then you put a telephone number there, then you are switching from digital to call, which may not be the best experience for many people. So in this particular one, you can see that when we started the chat bot journey, we had to start with about 1700 unique customers using the chat bot in the first month. And after about a year that almost travelled to 4,300. And we could see almost like for drop in the core traffic as well.

So once again, we weren’t actively promoting it was there as an additional way of handling customer queries and actually the customers like that and didn’t feel the need to call us up. So again, customer experience benefit and an saving. And the final case study is around knowledge management. So I think I mentioned before that one of the key things that comes up time and time again when you’re looking at customer satisfaction is knowledgeable agents. People want to call us up and get through to someone who can handle their queries, hopefully one and done and get the information that they need. So that’s one of the ways that we’ve implemented knowledge management. So we started off by putting up a number of articles into the knowledge management system that we thought would definitely answer some of the common customer queries.

And we also gave the mechanism for our agents to provide feedback on the knowledge management system. Was this a good article, was it helpful? What else would you need? And that was incredibly useful feedback and allowed us to really customised that information going onto knowledge management system to make it even more accessible, easy to use and more useful for our call centre agents. So what was the impact there? Over 18 months, the content library grew from about 70 articles to 250 and the page views went from 900 a month to over 3000. And over that period we can actually track that our average handle times again on our busiest product line, this Child Trust Fund that went down by 8%. And that was driven largely by the fact that the call centre agent had the information at their fingertips. So any hold time could be reduced or any handoffs to others were reduced as well. So that made the call shorter and more productive for both the customer and the call centre agent.

So hopefully you can see a number of examples there, real life examples where we use the AI technology to deal with a particular problem and you can see there that it delivered tangible customer experience and efficiency savings. So just by way of just summarising all of that, I’ll probably just go to the boxes, the stars there. I think using those technologies for me, for our customer operations, it split to less errors, higher quality scores, faster turnaround times. One of the things I didn’t mention is that using the chat box, we use the functionality there, which was digital uploads. So rather than customers sending in their paper copies of, for example, identity documentations, they could just take photographs and digitally uploaded and come straight to us. So a number of days in the mail down to a few minutes by digital channels. We were able to reduce the volume of incoming calls handled by chat bots. It gave customers more choice.

So one other example that we use chat bots for was the ability of customers to do a change of name change, of address, notification of death, all done via the chat bot. So no need to call us up. And of course by using a chat bot you can do that 24/7 anytime of the day, seven days a week. Using non-knowledge management. As I said, less hold time and a higher first contact resolution which was delivered by those more knowledgeable call centre agents. And probably one of the most important things for me is that we are able to do all of these things, deliver efficiencies to the customer operation whilst maintaining our high customer experience measures and customer satisfaction scores. In terms of the overall efficiency benefits, it’s a bit of a complicated graph. I quite like this one. Again for our main product, the Child Trust Fund over the last three years with a whole combination of the things I’ve just mentioned and a few other things, your usual continuous improvement lean techniques.

What this shows is in, sorry, in black is over that period. The actual activity relating to that book of business has gone up over time. But over that period, the actual manual effort to handle that increasing level of activity, which is the red light, has actually gone down. And we’ve measured that by the number of calls coming in the average handle times and also the effort that we see in the back office operations that we can measure via our workflow management system. And that is a really great measure of increasing efficiency over time. And most of that has been generated via AI technologies. And just to put that in context, that’s probably delivers around about 30 people’s worth of efforts of saving. So 30 full-time equivalent savings over that period of time. So quite substantial for an operation of our size.

So finally, just some key takeaways from me about our experience, which hopefully might resonate with you, but maybe some lookouts for you as well if you haven’t already started your journey. I think the big one for me, the top one is this should be considered business-led initiative, not an IT-led one. It’s really where you can get the biggest benefits, it’s in those business areas. So those business areas that are closest to the customer knows the issues. And also the great thing about these technologies, as I mentioned before, it’s a low code technology. You don’t need experienced coders to operate it. So that business-led lens means that you are really focusing on the issues, you’re focusing on the business benefits, and you’re looking at how that technology supports you to deliver those. I think on anything like this where it’s quite new to you or to the company, it just does help to get some expert help to get you started, to make sure you don’t make some obvious mistakes, and to get you started quickly. Select the best of breed vendors.

Just be wary about companies that can deliver. They say they can deliver everything and they’re expert in all areas, some maybe, most aren’t. So just spend your time talking to a lot of vendors and make sure that they not only do they have the right technology, but they have the right support structures and sometimes the right cultural fit. Identify those use cases and build a business case. So if you’re going to implement something, how can I measure whether it’s been successful or not? You saw some of the measures that we use, which is how does it reduce average handle times, how does it reduce volumes? And so what are the efficiency savings there? But also how do we measure the customer reaction to this in terms of lower complaints and better customer satisfaction times lower turnaround times.

Start small, test and learn then scale up. So really use an agile approach there. You won’t get it right first time. And the nature of these machines are, they’re learning machines so they won’t get it right first time either. So you do need to train these technologies. Every company is different. So don’t think that it’s going to happen instantly. These things do take time to teach so that they can actually be very tailored to your business. And that’s the sixth point, don’t rush. It takes time to teach the machines. Be safe, very important one. So you really need to adhere to a responsible AI framework. So you see a lot of information out there where they’re talking about generative AI that is printed by us. It can hallucinate, give the wrong information. So it’s really important that, as part of your risk management framework, you do focus on those particular risks and you mitigate against those.

One of the things I should have mentioned before, involve your cyber team up front. We lift that a little bit late to our cost. And that did introduce quite a few months delay. When I say delay the type of cyber team, were very interested in what we are doing and quite naturally if data is flowing in and out of the organisation to other third parties, they need to be absolutely satisfied. That’s being done absolutely securely. So my advice is get them involved up front and that will probably save you time throughout the project. Build internal capability. So I mentioned about the local team where we looked at some of our really high calibre people in a customer service operation who are that aptitude for working with this technology. Great development opportunity for them and great for us in terms of building that internal capability and not having to rely on too heavily on third parties.

And then the last one is, although we are just dipping our toe into it now, get ready for generative AI. It’s here already. It’s big, it’s getting bigger and I think the future is all about that generative AI. You can see that there’ll be lots of benefits coming out of that in the future, but it does need to be done in a sensible measured way using a responsible AI framework. So get ready for it. One way of getting ready for it is to get good at the more traditional AI technologies such as the ones I’ve just mentioned, speech analytics, chat bots, and knowledge management. So I hope you found that useful and it’s giving a few little hints and tips and some ways of measuring how you go about assessing the benefits of doing these and a few of those case studies, which I’m sure will probably be relevant in quite a few of your organisations. So thank you for listening. Over to you, Ben.

Ben Telfer:

Thank you very much, Steve. I didn’t mention at the start of this webinar, but we have so many joining us today and I think that was because of the promise of some practical examples and some use cases as well as some teachable insights that you’ve shared and it didn’t disappoint. So thank you for everything you’ve shared today. We do have a number of questions. I don’t think we’re going to get to all of them. We’ve probably only got another five minutes, but if anybody has any questions, please do send them in afterwards and I’ll pass them on to Steve. I’m hoping Steve, you’ll be very willing to connect with other members and share a bit more on your experience so far.

Steve Ferrari:

Absolutely.

Ben Telfer:

One of the questions, you touched on it just at that end around gen AI, and I think this is the big thing that a lot of people, AI is not a new thing, but obviously the rhetoric in the last and the narrative over the last year, two years has been around gen AI. Are you able to share some of the things that you are investigating with Gen AI at the moment at OneFamily?

Steve Ferrari:

Yeah, well firstly, there’s probably quite a few of your suppliers already or third parties that you deal with who do use Gen AI. So for example in IT for cybersecurity, I know a lot of those companies use Gen AI already. So although you might not be using it directly, I’m sure some of your suppliers might be. For us, some of the things that I’m looking at, take for example our call centre. So at the end of every call we have the wrap time and our agent will make some notes of the telephone call. So gen AI, certain applications will allow you to do auto-summarization. So rather than the agent do that at the end of the call, the call will be summarised and all the key points picked up and then it all just requires the agent just to do a quick check. So always having the human in the loop and then press submit and then at least we have some really good rich notes of the customer conversation, which can be used when we’re referring to listening to calls again.

But also that you can pump straight into a customer relationship management system. And using that similar sort of technology, we have a small financial advice arm and we have looked at certain companies that provide that whereby let’s say you’ve got a client conversation with one of our advisors, it lasts an hour. Right at the end of the call if it’s recorded via teams or some other recording technology, then it will just summarise the call immediately and produce things like suitability letters, summary emails to the client. So there’s a number of those applications that we’re looking at already. But that’s just scratching the surface. There’s so much more.

Ben Telfer:

Thank you, Steve. Thinking about the values of a mutual and cooperative insurance company, how do you feel those align with the use of AI? And also is there anything about your mutual status and the idea that OneFamily is very much a customer-centric purpose led insurer that has helped set you up for success in what you’ve done with AI so far?

Steve Ferrari:

Yeah, well as I mentioned before, I’m not going to do anything that’s going to compromise our customer experience. So our customers are members, they’re number one. So for me, whenever we implement any of these technologies, it can’t be at the expense of that customer member experience. What I’d like to think is that when we do this, it’s a win-win. So we improve the customer experience as well as improving efficiency. And we have to look after, protect and grow our member assets and becoming more efficient allows us to do so. For me, it’s absolutely aligned with our mutual ethos because it allows us to become more efficient so that we remain a sustainable strong business and able to deliver benefits to our members through investments in new products and so on, and whilst at the same time delivering any and improving customer experience.

Ben Telfer:

And just to expand on that, I’m assuming you connect with peers across the UK financial service landscape. Are they taking a similar approach in terms of it being a customer-led initiative or for those, is it more technology or data-led projects?

Steve Ferrari:

I think it varies. I think most people are genuine when they want to deliver this technology to improve the customer benefit. I’m sure some organisations go through it from starting from the cost savings potential, but ultimately a good service costs less. So I think when you start off with that customer experience as the lead, if you improve the experience, you provide more choice, you reduce the turnaround times, you improve the processes, which makes it easier for the customer that has that knock-on benefit of delivering efficiency savings as well.

I think some comments I’ve had just recently going to a number of conferences when I’ve explained the experience that we’ve had and what we’ve been able to do over the last year or so, the last couple of years, some people have said, “Well how come you’ve done it so fast?” Because IT are always so slow on this thing. I just said because business-led, not IT-led. I know we are going through a platform rationalisation programme at the moment. I know our IT colleagues are absolutely snowed under. So perhaps for them this be the highest priority. So if it’s more business-led, I think, you’ve got standard a better chance of getting it done more quickly.

Ben Telfer:

Thank you, Steve, I think we’ve got time for one more and it’s just about regulation. Obviously you’ve touched on a little bit there around the idea of responsible AI and some of the ethical considerations. I know we’ve just recently seen an EU AI Insurance Act or AI Act, I should say. Have you got any idea where the regulation may go in both the UK and more generally worldwide? Because it seems like it’s going to be going to some way of regulating the use of AI in financial services?

Steve Ferrari:

Yeah, that’s a big question. So not quite sure. I don’t think I know the answer to that one, but I think what’s clear is, especially with the proliferation of gen AI in the last year. You do read some pretty awful stories of how it can be used by bad actors. So very much welcome the right regulation around AI so that we can try and stop those kind of behaviours and those risks. But it’s always a trade-off, isn’t it? If you go too far then you might stifle innovation and creativity, but if you don’t go far enough, then it opens the door to potentially bad things happening. So it’s important that the right balance is struck. For us, it’s our responsible AI framework. We don’t have it as a separate thing. It’s part of our overall risk management framework, which is as it should be. And really we’re not going to implement anything where there’s a risk to our customer’s data or bias or the technology giving out bad information.

So for us, certainly in the initial period, if we do, not if, but when we go into gen AI, it’s going to be very much keep it internal first, keep the human in the loop, learn, get the experience before you roll it out more widely. So really taking a measured approach to it.

Ben Telfer:

Super. Thank you very much, Steve, for sharing a quite good response to a very expansive question. I think in the same way that you know gen AI is coming, we know regulation is coming, so again, it’s just setting up the frameworks to be compliant to that. We will end the webinar there. Thank you everybody so much for joining us today, and thank you to Steve for sharing such detail of what’s going on at OneFamily and answering some questions there. As I said before, if you have any other questions for Steve or OneFamily’s experience, please do contact myself or get in touch with ICMIF and I’ll happily pass those on to Steve. We’ve had a lot of conversations with members recently on the topic of AI and you’ll see on our ICMIF knowledge, Hub, there’s numerous assets on there in terms of thought leadership articles and other case studies around what other insurers and mutual cooperative insurers are doing around AI.

ICMIF itself is also just about to finish a survey of our members in terms of their approaches to AI and we are hoping to publish a short report on that very soon, which will include some further use cases and case studies from our members in terms of their approach to AI. So look out for that very, very soon. And just finally, again, thank you for joining today’s ICMIF webinar. We have more than 130 other webinars that are available on demand to all members, and you can access that via that link on screen. So just final finally, a big thank you to Steve again for sharing OneFamily’s experiences. Thank you for joining us today and see you on another ICMIF webinar, ICMIF Virtual event soon. Thank you, everybody. Enjoy the rest of your day.