Subscribe
Share
Share
Embed
Artificial intelligence is changing the way real work gets done. But big ideas don’t drive change. People do.
The ones who roll up their sleeves, modernize data, and bring AI to life where it matters most. In the workflow.
This is for them. For you. The visionaries. The innovators. The leaders turning potential into performance and pushing their organizations forward.
Everyone’s talking about the promise of AI and what it can do. On this show, we’re talking about making it real.
Learn from the experts who are driving it forward and walk away with everything you need to bring AI to life in your organization.
Tim Rafton (00:00):
The AI's superpower is this unstructured data bit. And so moving away from large multi-year programs where you replace entire platforms may still need to occur in some cases, but in the meantime, you don't need to wait for that. You can really use the AI and use that superpower to get really strong quick momentum, solve problems in short birth.
Shirley Macbeth (00:26):
You're listening to Make It Real, brought to you by EXL. I'm your host, Shirley Macbeth, and on this show we're exploring how artificial intelligence is reshaping workflows, industries, and the way real work gets done. And yes, we're going to make it real. When it comes to transforming workflows with AI, the real world has a habit of getting in the way. Often it's easier said than done. You've got legacy technology, legacy data, resistance to change, and all of that can make transformation difficult. That's why I'm very excited for this episode. We're going to hear from an amazing leader who has tackled all of this and is successfully driving and has a vision of a roadmap around innovation with AI in his organization, and he's going to share his lessons with us today. So joining me today is Tim Rafton from IAG. He's the executive general manager who oversees claims and operations for their personal and commercial business lines.
(01:24):
And for more than a decade, he's been focused on pushing digital boundaries within a very highly regulated and obviously risk-averse industry of insurance. So welcome, Tim. Thanks for joining me today.
Tim Rafton (01:35):
Thanks, Shirley. Thanks for the introduction and great to be here. Great to
Shirley Macbeth (01:38):
Have you. Well, we'll jump right in. So EXL and IAG have been working together for about 10 years now. And so Tim, what trends and challenges have you observed during that time when it comes to first digital and now AI adoption?
Tim Rafton (01:51):
Yeah, thanks, Shelly. And look, probably the themes I think have probably been similar. I think the tackling of them has certainly evolved. And with this recent development of AI and the speed at which the adoption of AI and putting it to use is a game changer. But I think to your question, the trend has really been about two things. One, how do you service a customer more effectively and efficiently? And how do you drive that efficiency through your business in a way that it's taking cost down, you can pass those efficiencies on through their premiums and the like. The challenge, of course, particularly for an organization like IAG and many insurers are in the same boat. It's a company that has a long history. It's been around the roots in some of the brands have been around well over 125 years. And over that time, there's been this collection of a lot of other insurance companies, consolidation and acquisition.
(02:41):
And so with that comes a lot of legacy systems, a lot of legacy processes and ways of working. And traditional digital and automation have really needed cleaner technology stacks, but also cleaner data to truly get automation. The game changer now for AI, as I learn more about it, is really AI superpower is its ability to deal with unstructured data. And so as we still pursue this cleaner stack of tech and data and the way we organize it, rather than waiting for those bigger transformation programs to come along, AI is really unlocking some of those efficiency and customer gains in the short term. So we don't have to wait for all that cleaning up to happen. We can actually solve some real world problems now and make the life of our people who are servicing customers a lot easier.
Shirley Macbeth (03:31):
Yeah. When we spoke in the past, you said there was some low hanging fruit in some ways that you can really maybe test AI and in terms of making some quick gains to advance the comfort level within your organization. Can you talk a little bit about where would you start?
Tim Rafton (03:47):
Where I would start is solve some actual problems for people. Rather than finding the thing that has the best business case, lean canvas numbers on it. Actually talk to your team, talk to your people, listen to the verbatim comments from customers, think about where are those friction points? Where is it that they get most frustrated or they lose a lot of time in getting to an outcome on a claim or any process. And so that would be my starting point firstly. The very first experiment we did was one of those real world problems for my team was, again, reflecting on our history and the breadth of the customers we service through intermediary, we have about 1,200 product disclosure statements or endorsements or contracts of some description, incredibly challenging for people to navigate and try and understand, as well as as we've moved more and more to work from home or hybrid work environment, our traditional learning models relied on peer-to-peer learning.
(04:44):
If I hear and listen and see what my more experienced colleagues do, we don't have that luxury anymore. And so the very first use case was to load all those PDSs into one spot, that in itself solved the problem. Previously, they were in multiple locations, and then we built a bit of an insurance brain around that using AI and large language models, and gave our teams some chat functionality. They were able to prompt the AI to help them not just find the right contract or product, but also find the right information depending on the particular question or problem they were trying to solve. So a fairly small use case and in the scheme of things isn't going to revolutionize the whole process, but it actually opened up my eyes certainly to just two things. One, how quickly we were able to do that and put it in the hands of our people.
(05:39):
And secondly, the appetite of our teams, no matter their demographic, tenure, years of experience, everyone has this huge appetite for AI. And I think they have this appetite for it because they can see it does make their life easier and it does beep things up.
Shirley Macbeth (05:54):
Absolutely. Well, I think that's very exciting. I think getting it into the hands of your employees and then learning from how they use it and to solve real world problems. I mean, not doing tech for tech's sake, it's to actually solve some real world problems. You've also recently deployed your first autonomous agent, and I was wondering if you could tell us a little bit more about the problems that it's solving.
Tim Rafton (06:16):
Yeah. So we went through a bit of a process of firstly just mapping out our processes and getting a proper end-to-end view of all the moving parts, all the decision points. We used a lot of the data that we have available to us and also overlaid things like customer surveys and complaints data to really get this clear picture of how the permutations of a claim as it goes through start to finish. And out of that, firstly, what we said to ourselves or had a good think about was, well, how do we first just make the process a bit more modern, a bit simpler, meet contemporary needs and expectations of consumers and customers and our people, by the way. And so that was a starting point. To your part of your question there, there's no point using AI to administer a dumb process. We first make it smart and then deploy the AI or the automation or the bot, whatever it is to administer an efficient one.
(07:17):
Where we then got to was starting to think about where in the value chain might we want to start. And off the back of the other experiment, I just touched on one of them, we have this huge appetite and a bit of a roadmap now around how do we get to some autonomous AI, eventually getting to more sophisticated, agentic orchestration. The part we started was at the very start of a claim process, if you like. And so we started to look at things like what happens at renewal or the inception of a product that maybe slows things down once we get to a claim, and we sorted some things out there around data ingestion and the way we capture information about customers. Then the actual claim lodgement itself. Now, this might be hard to believe, but still 30 to 40% of our claims are lodged via an email or a web form.
(08:04):
And we have digital solutions and other channel, but still email is relied upon by many of our customers and intermediary. And as we started to embark on figuring out how we would deploy an automated autonomous agent in that area, interestingly in that pilot, one of the examples we got of a claim lodgement was a customer as a farmer in one of our rural parts of the country, and they sent us two photos. One photo was of their certificate of insurance for a piece of farm equipment, and the other was a napkin with a handwritten note on it that said, "Lodge my claim." And that was the extent of the information and the data we got. And so as you sit and try and think about that, you either need to change the behavior of the customer or what we chose to do was think about, well, how do we now start to put some things in place where the AI, the same as a human, would otherwise need to start to navigate that process with the customer and guide them through it and nudge them onto digital channels.
(09:03):
And so this autonomous agent now is able to take that first email, take from the photos that we got, the data that is relevant to the claim lodgement, the vehicle details, the customer's name, their address, their phone number, all those things, and then put it into web form, identify the parts that are missing, and then go back to the customer and say, "Thank you for providing that information. Can you just give us these other bits that we need?" And then once they send that back via our web form, the AI agent will take that and it'll continue the lodgement of the claim. So a very, again, a smallish step in the context of the whole end-to-end, but incredibly exciting because what that does is unlocks a whole bunch of capacity for our people, takes them away from admin and allows us to now start thinking about how do we focus them more now on the softer skills that are required in the management of those types of claims or other claims that need a little bit more hands-on attention.
Shirley Macbeth (10:01):
That's such a great example. And you call it a small example, but I think it's a really big one because when you think about all the pieces that are involved with that and the napkin is such a great example of unstructured data and you're not going to change the behavior if your client is trying to engage with you and that's the way they're doing it, you're meeting them where they are. You're getting that into the workflow, you're then prompting for more information, but you're digitizing and bringing that. So I think that's such an exciting example. You started by saying that AI's superpower is really around the unstructured data, and I think that's such an interesting example. Is there other lessons you've done a lot around data that we had talked about, you're undergoing some data migration, legacy systems. How does this whole AI world play into that as far as getting your data ready for AI?
Tim Rafton (10:52):
Well, I think as the data transformation itself occurs within our organization and we realize the benefit of that, it'll make the availability of use cases far more accessible and probably far more powerful. Things like the orchestration of an entire value chain of a claim probably become possible. I think in the meantime though, what the work we're doing now is allowing us to do is take that unstructured data and put it into a more structured format. So it's almost solving part of the problem for us in some way. But I think secondly, it's also giving us some different insights into our business, as in we are now starting to think about how do we prompt the AI or give it the business rules that allow us to pop claims out to be escalated to a certain skillset or a kind of person that has the right ability to deal with a certain set of circumstances.
(11:50):
In the past, would rely on people remembering or knowing or necessarily having all the knowledge of what to do in every single set of circumstances. And so I think it's not just about having the data structure, it's about how do you use the AI in a way to really help you understand the risks, the inherent risks in your business and make sure that they are flagged or escalated and they go to the right people or triaging or some other process that needs to be applied to really get the right outcome for the customer and that client.
Shirley Macbeth (12:22):
It's very exciting from a customer experience point of view that you're able to triage and you're able to really help your agents and others that are on the phone, your people agents really handle the tough cases and solve more challenging problems. So I think it's really, like you said, it's solving broader issues and helping you provide that level of customer service that's even more advanced. That's really exciting. I'm curious about the journey that you went through. You said you've got these small use cases, but really you've set a very aggressive and broad vision for IAG with where you want to be over the next several years and how you would get there using AI and other technologies. And that requires a lot of buy-in. I would imagine from various stakeholders, executive team and others, you're in a regulated industry. Tell me about some of the considerations to keep in mind for that.
Tim Rafton (13:15):
Yeah, and the stakeholder part I think is critical. For me, I'm lucky enough to have some peers that are also just as passionate and can see the potential and the benefit of this type of technology. And so we've set up a really strong trio of co-sponsorship within our team. I think that's really critically important because my decades in the insurance industry is in claims. I'm not a tech person, so there's no way I can do this or bring this to life on my own. So I think that foundationally really critical. It makes sure you get the right sponsorship buy-in and everyone's really clear and aligned on what you're trying to do. I think the second bit then is, and you touched on is being really ambitious in the goal that you want to set, and it needs to feel achievable, but it also needs to feel a little bit uncomfortable too, in my view, not just as it relates to AI.
(14:07):
Any transformation or evolution of your business I think needs to have some hefty goals. We've set a goal that says we want to design a process and an AI capability that enables 100% straight through processing of a claim. Now in that, we'll make some decisions about certain things that we don't want to just have an AI agent autonomously deal with. We'll want to pop some things out and make sure we have a human in the loop, but that's a conscious decision. The point here is we want to design it to be 100%. If you use Waymo as a great example, the autonomous car, they've designed it to be driven without a steering wheel, without a human behind there so they don't need a steering wheel. But then to make people feel comfortable in the early days, they've put a steering wheel in there and put a person behind the wheel.
(14:52):
They don't need the steering wheel. It can be 100%. And that's what I mean by a hundred percent designed for straight through processing. And I think that's really important. A lot of people are already saying that's not possible, but we're really holding the tension on that straight through to make sure we get the design element perfectly right. And then I think that the other part here is to start to work backwards then on building momentum. And we're using that term a lot, but really not just transforming the use of AI. We're transforming the way we deploy technology into our organization and moving away from perhaps would've otherwise been quarterly planning cycles to now fortnightly releases and fortnightly optimization of certain elements, which in itself is really, really critical. We can just continue to build this momentum and gradually over time add more and more feature capability and autonomy as the agents get more sophisticated.
Shirley Macbeth (15:50):
I think it's tremendously exciting. And you say momentum. I think of what, to me, it's translating to speed too, this constant innovation and forward momentum towards a larger goal. When we've talked in the past, you've called it a step change. You're not planning for incremental for 30%, you're planning for a hundred. And when you're doing that straight through processing, with that as a goal, your mind is, I think, opened up. And maybe that's part of what you've got this alignment, you said with the trio of executives that you work with, you're aligned towards something that is very transformational for the business.
Tim Rafton (16:24):
That's a really good point. And you're spot on. That was a discussion we had very early on that if we aim to incrementally get better, we may be 30% more efficient or we may take 30% of the steps out in the process, which would be great most people would be happy with. But actually aiming for a hundred, even if we don't get to 100%, we might end up at somewhere like 60 or 70% better or 60 or 70% less time spent on a claim, which is far better than the incremental sort of mindset. So the goal itself is really about trying to drive a mindset and really get people galvanized around the goal that we have. And the goal that we have here, as I said at the start, is to make the experience from a customer point of view better. And we know to do that, we need to give time back to our people.
(17:16):
We need to take them away from the admin, and we need to allow them to deploy those softer skills that only a human can really do in a sophisticated way.
Shirley Macbeth (17:24):
I think that's great. Well, I think I want to end where we started with, which is around the people who are actually driving some of this transformation and the people that understand these problems that are in the field that you're dealing with with real world examples. And so we talked a little bit about the executive buy-in, but let's go back and maybe close with the people that you're learning from and are building some of this innovation. You said initially that there's actually a lot of excitement around that. So how did you really bring them along and then enable that excitement to build out some of these solutions that you're talking about now?
Tim Rafton (18:04):
Firstly, it was about solving actual problems that people are telling us get in the way of doing their job really, really well. I think in doing that and proving that out, you'll get buy-in immediately. And we've had no shortage of appetite for more. In fact, off the back of that, the backlog we now have of, can it do this? Can it do that? What about this use case? Probably near impossible to get to all of that in a reasonable amount of time. So the appetite is definitely there. I think the second bit I would say is the AI's superpower is this unstructured data bit. And so moving away from large multi-year programs where you replace entire platforms may still need to occur in some cases, but in the meantime, you don't need to wait for that. You can really use the AI and use that superpower to get really strong quick momentum, solve problems in short bursts.
(18:59):
And it is only incremental, but over time when you look back, those increments will add up to having solved quite a bit of the frustration and the puzzle that often can be unstructured data. And then I think the third bit is figure out how to meet the customer where they are. I think that was the language you used, Shirley, which I think is a great way to put it. And for us, in our commercial business, we are dealing with products and brands and agents across a breadth of different needs. Everything from your personal motor vehicle or home through to farms and homesteads and farming equipment to workers' compensation and professional indemnity and liability. And so all of those customers, all of those products, they're all different and not everyone's understanding is the same. So rather than trying to use the AI to get funnel people into working the way you want them to work, really spend some time thinking about how to use the AI to go and meet the customer, where they are and where their needs are.
(20:04):
And then in the background, figure out that funnel without having to make them step through too many hoops. I love that. And we've seen that with this first deployment. It's a really good way to think about it.
Shirley Macbeth (20:17):
Amazing. Well, Tim, you've given us such great advice and such insight into the challenges that you're solving using AI at IAG, and it's amazing to see the progress and learn from what you've done. So thank you for spending the time. To me, there are several things that stood out. So I'm just going to try to summarize them quickly. Number one, you said use AI and think about AI to solve actual business problems. So you're not doing tech for tech, you're actually leveraging technology to solve actual business problems. That's number one that stood out to me. The second is really you've set ambitious goals for what you are doing, but you're not waiting forever. You're taking steps to get momentum and speed in these bursts that you were talking about. And so you're solving these, what you call small, but they over time add up to some very big transformation that you're doing across the company.
(21:07):
So that was number two for me. Number three, I love what you said about meeting the customer where they are. That's tremendous and leveraging technology to solve business problems rather than having the client have to conform to what you're doing you're solving with that great example that you did with the napkin. And number four, I think there's a lot around executive buy-in and working as a team that has really helped to shape the organization around what you're trying to do. So Tim, did I do okay as far as summarization?
Tim Rafton (21:37):
Great summary. Very good summary. Yeah, that's exactly right. Okay.
Shirley Macbeth (21:40):
Well, Tim, thank you again so much for spending the time with us. I think our listeners are going to really love to learn from what you've achieved and the vision that you set forth. So thank you very much for spending the time.
Tim Rafton (21:53):
Thanks, Shirley. It was great talking to you.
Shirley Macbeth (21:57):
Thanks for listening to Make It Real. We hope today's conversation gave you ideas, insights, and inspiration to help bring AI to life in your organization. Remember, big ideas don't drive change, people do. Keep learning, keep experimenting, and keep embedding AI where it matters most. Follow along so you never miss an episode.