Taming the messy middle with Nigel Walsh on intelligent insurance workflows and AI transformation

Welcome to Unstructured Unlocked, a podcast where listeners discover how insurers are entering the decision era, utilizing artificial intelligence to refine their decision-making processes, boost underwriting profitability, and achieve premium growth. Hey, everybody. Welcome to a new episode of Unstructured Unlocked. I'm your co-host, Michelle Govea. And I'm your co-host, Tom Wild. Well, Michelle, today we have a terrific guest, someone who's...

traveled a lot through this industry and has a really unique perspective. Nigel Walsh is the global head of insurance for ServiceNow. I think this represents a really interesting conversation because perhaps historically, ServiceNow. would not have been seen as sort of sitting in the insurtech landscape. So maybe a relatively newer reality for ServiceNow. But certainly, Nigel, we're fascinated to hear about that journey. I think the topic today is...

is all about sort of core platforms and then that messy middle challenge and how the growth of AI agents will impact that in the future. So welcome to the podcast, Matt Joel. Welcome. Thank you so much. Very much looking forward to it. dare I say, one of my favorite topics. I've been talking about this for what feels like 20 plus years right now. So I'm not sure if it's going around in full circle or not, but yeah, let's get into it. Perfect.

Well, maybe start with a little of your background, a little of those travels. You certainly had a successful journey at Google and now ServiceNow. I'd love to hear your perspective broadly on the industry. And then how did you... end up at service now here running a mobile insurance company. Yeah, as you say, I was at Google Cloud in New York, having an absolute blast. I've been there about four years. It was a phenomenal organization.

anyone know. It was the first organization in my entire life that I didn't have to tell my mom or any of my friends what they did because everyone has a perspective of Google, of course. Now, what I actually do in my day job, people still don't know, even my wife of 25 years. But there we are. But Google for four years, prior to that, as a partner at Deloitte in the UK, running Global Future of Insurance and Global InsurTech. And prior to that, Capgemini doing large-scale transformation of...

Usually PNC carriers with the likes of Guidewire and Salesforce and Duck Creek and Pega and a bunch of other core applications that are nascent, I think, across the globe for almost every carrier. The ServiceNow gig appeared in a very intriguing and interesting way. To your point, when they called back in July, August last year, I said, I know who you guys are. I just haven't seen you in insurance before.

well, for a very long time, or at least in the area of insurance that we operate in, in terms of line of business and much more. And after you unpick the covers and see what their 7,500 plus customers are up to and the fact that we have over 500 insurers using the platform today, you then start to see how their move to verticalization from...

doing what people know as traditionally as ticketing or ITSM and everything else, and how that's expanded over the years to HR, governance risk and compliance. And now there's a real focus under Holly, who leads the global strategic industries team. into key industries, financial services, retail, telco, manufacturing, et cetera, et cetera, and then how the platform can be applied to solve, in essence, what Fred Luddy set out to do years ago when he...

created the platform, which was move work around an organization, the first thing he chose just happened to be IT ticketing. If you think about claims or underwriting, in essence, all we're doing is moving work into a business, around a business, and then back out somewhere. We've got ourselves into a really interesting and intriguing spot that I think as we go through this, you'll go, you know, I didn't know you guys did that.

And that's been my favorite phrase from customers over the last six months is we get into a conversation and the next minute I go, all I ever hear is, wow, we didn't know you did that. It's always an interesting journey to...

career-wise. So I love hearing those stories in the beginning. I'll jump right into it, Nigel. So Tom mentioned at the start, you know, talking a little bit about core systems and AI and transformation. We've spent a lot of conversations or we've had a lot of conversations talking about core system transformation and data transformation, how that's been almost like a building block or a starting point for a lot of carriers over the past, you know.

I'll call it 15 to 20 years where you had some carriers doing core system, big transformation projects early on with what are now known to be the common names and incumbents. Curious your thoughts on from... How does AI change that core platform modernization discussion? Does it eliminate it? Does it change it? Is there another wave of transformation that has to happen where even those new... new core systems become obsolete or legacy within the next two to three years?

Wow, that's a huge question. I always start off with the falls, right? We can spend the next 30 minutes just talking about this and still not be done. In short, is there going to be another wave of transformation? Yes. Dare I say, I'm old enough to, my career started out on deck back 6420s and 3210s doing mainframe and what's not in a very old world. You fast forward a decade or so and-

Folks like Guidewire truly have created a phenomenal modern-day platform for P&C carriers worldwide. DXC on the L&A side, for example. There's a whole host of folks that have done or created like... core foundations for the modern insurer. But if you look across the world, across the various different landscapes, the speed at which each of those or any of those carriers move is dependent on their core platform that was built in many cases for yesteryear.

And as they migrate onto modern day platforms, like the ones we just mentioned, I think there's an opportunity to work out what the modern day customer seeks. Because if we go back 20 years, one of my biggest bugbears in this industry is we are... policy-centric, not customer-centric.

And that creates an issue. Like when you call your health insurer today or your auto insurer, the first thing you're going to get is what's your policy number? Not hi, Michelle, how can we help you? Or whatever it might be. That's driven me mad for a long time. So then came the whole data conversation about actually coming up strapped out, if you remember that two-speed IT from Gartner years ago, or that multi-tier, you know, three-tier architecture between application.

data and interaction and so much more. We're starting to see a shift back to that, where you've got either a system of record that might cover, hey, here's all my core data or where everything is stored and processed. And that might be in a core platform or it might be in something like a Mongo, a Databricks, Snowflake. lake or whatever else. I think lots of carriers have tried to abstract that away so they can use it in multiple different ways.

You've then got that core, how do we go do the processing piece? And then you've got the next layer up, which is workflow, interaction, engagement, and so much more. And folks like ourselves or Salesforce and lots of the low-code, no-code platform. have done a good job at making that more flexible. I've long said, the closer you get to the core, the harder it is to transform or do things that are unique and differentiated.

But the further away you go, if that's a concentric circle moving outwards and the customer is at the very edge of that circle.

with a contact center or face-to-face contact, there's loads and loads you can do with less disruption and more agility. So I think that's where AI started, all those use cases for generative AI. And since then... agentic have moved from the edges and they're getting closer and closer to the core, which has got a bigger impact and more value ultimately for both the customer and carrier.

A long answer, I'm afraid, but it's, yes, it will impact. Yes, we've got another wave of change. And with everyone's desire to squeeze cost or take out inefficiencies. I think we're going to see the agentic AI and AI era really challenge that system of record, the data capability, and system of engagement.

That's a good segue to this concept of sort of the messy middle. I think historically in insurance, driven by the fact that you do have... companies that live in multiple eras simultaneously, right? You've got green screens, especially in life. You've got green screens. You've got, you know, kind of more modern applications at either end of those spectrums. And then how do you...

stitch all that together made it even more complicated by the fact that insurance has grown a lot by M&A, right? So you end up with two and three and four and five versions of all possible platform types. You know, I think the first approach to solving this was just offshore a ton of manual work, right? And just brute force this. And then you had the RPA era maybe, you know, 10, 15 years ago, which was an attempt to sort of...

create flexible piping between these things. Now we have the rise of AI and agents. Does that allow customers to not have to do the giant re-platforming anymore? Does that unlock some of that or are they still there? I mean, how do they keep up with modernization knowing that these can be $100 million plus multi-year pre-platforming projects, at least historically, if you look at historical data?

Yeah, I think you're right. There are $100 million plus and then some, depending on the size or scale of the organization. I think you're right to break it down between LNA and P&C. LNA transformation cycles are roughly every 40 years. P&C are every seven. Our PNC transformation is easy because you can migrate on manure on a 12 to 18 month basis that are relatively straightforward.

And I say that with a pinch of salt, but on a rather straightforward, clear-ish perspective, with an L&A book, it's much more complicated and difficult, specifically if you're locked into 20, 30, 40 years of legacy, whether it's open or closed book or whatever else it might be. I do think those modernization programs are still underway. We still see quite a lot of those worldwide.

I think people are now questioning my, do I want to lock into a monolith or can I break this out into more of an agile capability in the same way that... I'm going to ingest data from not just one source, but maybe 100 sources. And I might have seven or eight sources for each line of business or each area or different countries or whatever else it might be. I think that...

And again, I might be biased or otherwise, but I think that workflow across the organization, that steel thread that shows process end to end is critical. What hangs off of that over time, I think will change. Like if we went back 20 years.

We'd say one of the things that hung off a workflow might have been facts, but today it's going to be voice conversations that are transcribed into data for us to understand and get sentiment out of. That wasn't possible 20 years ago, but the workflow process. is in essence very similar to what we had 20 years ago. We might have changed it or refined it or whatever else it might be. So I think there's going to be a continuous slow move.

of the modernization of the legacy, I'm not quite clear in my head about what that new to be state might be. I do think it's going to be more agile. What I mean by that is I think people want the optionality to be able to swap things in and out efficiently. If my telephony provider changes from Cisco to Genesis to Nice to...

five nines or whoever else it may be, I want to be able to plug all those things together. So the ability to plug those things in and out and switch easily is really important. Yeah, of course, the technology piece there is just one angle. And if you look at my old shop at Google or any of the hyperscalers, they all said you can move data free of charge and get out of the hyperscalers.

But that's just part of the problem, right? You might be able to get out free, but you've also got to then work out what the business change element is. And that's true for any of these things we're doing around process as well. Najee, you mentioned there's the technological...

challenge or the solve that AI brings. Curious on your thoughts on the control piece. So with new data, with new types of... elements that come into the process. There's a new world of what controls have to be in place. I'm not even talking from like a regulatory standpoint. I'm just talking about internally. How do you manage the flow of data, the usage of that data?

monitoring for biases that come out because you've got more of a automatic processing that goes on? How do you think about, or how should carriers think about the control element as it? as it aligns with the technological shift that they're trying to implement. Yeah, it's a great point. And again, it's another one I've been noodling for probably a decade plus is application sprawl has been at the...

heart of my career for probably 15 plus years as we, you know, to Tom's point, we layer in new things. We want to write a new piece of business. We'll just add something into the estate and whatever else. And all of a sudden you've got... lots of spokes and hubs. It can become quite messy, back to your point previously, about the messy middle. I think the same is starting to happen now in the world of AI.

where we've been through the phase maybe 18, 24 months back where we've had POC purgatory, as many have called it, as in you might have a chat GPT or Anthropic or Gemini or choose your own model or even your own stuff that you've done internally or agents from third-party software providers.

And they're all disconnected doing their own thing without any clear view or way of actually understanding or proving to either internal governance or external regulator what's been done at what point by who and when. And I think that's actually going to be...

one of the next battlegrounds. I was very pleased to see both the launch of the A2A framework, the agent-to-agent framework from Google a while back. And they've now since donated that to the open source community. And we, for example, we've got...

by pure coincidence, Deloitte, Google, and ServiceNow announced a whole load of agents that conform to the A2A framework, that you can see all these things worked all the way through, as you just pointed out. So I think having that framework in place Now, having learned from the past issues around data is actually going to help us create a better future rather than trying to solve for it after like we're doing right now with lots of the challenges in insurance.

Really interested to talk about that. I want to get there for sure. Before we go there, one of the other interesting realities of global commercial and specialty that I've observed is no one is really... driven sort of global platform decisions, right? They tend to be quite geographic in the way they make their platforming decisions, you know, between, let's say, the big regions, North America, Europe, and Asia, and even UK in Europe, you'll see.

the same business making different platform decisions. Does that surprise you relative to other industries you've observed? I'm curious if ServiceNow has a particular, you know, different approach to that. I mean, you know, in terms of how ServiceNow has observed customers. buying and deploying their platform. But that sticks out to me as interesting here, the very diverse decisions, you know, the same company will make across regions. It's really interesting.

And this is, again, another one that's been hanging around my mind rent-free for quite a few years. I don't think it's a case of technological inability. I think lots of technologies can do this. I think it comes down to corporate governance. dare I say, even politics in some cases, and speed.

And what I mean by that is I remember working with a carrier many, many years ago that could have deployed a core platform for admin, the same core platform for admin across multiple countries and multiple lines of business. But what they decided instead was because those lines of business moved to different speeds, they had one for large complex, one for MLC, one for SME.

And then optionality to do different things in different countries. So I see loads of similarities between, let's say, US, UK. But then as you go into Europe and APAC, I see slight differences around. be it regulation or whatever else it might be. I've seen lots of European platforms be very successful in Europe, but not expand out that way. And when I say Europe, I mean Central Europe as opposed to UK and everything else.

I don't believe it's a technological limitation. I think ultimately it comes down to speed of line of business and corporate governance to say, if we do this in UK and US and Germany and, and, and. then we can only ever move at the speed of the whole thing, rather than waiting for your individual piece. And over the years, I've designed things like, let's have a core platform with a country there, then a local there, then a regional there.

they just it can be hard it can be hard to do so i think it comes down to agility at the end of the day and the speed at which your business wants to move yeah it's an interesting challenge for Vendor partners like Indico, perhaps like ServiceNow, because you end up having many commercial engagements across even the same organization and supporting them uniquely. So it's an interesting challenge.

Yeah. And I also remember working in a telco long before I jumped into insurance and their biggest challenge was integrating two gargantuan or was it more than two actually gargantuan. ledgers together that were actually the same platform, but just didn't talk to each other. So again, that goes back to our earlier point about how I think this...

data might be abstracted from the core system in this instance. Because if you've got your individual workflow or process layer per country, per line of business, but your data is a single record across the globe, again, I've seen a few carriers over the years. years do that where they've got a common standard for customer. Imagine what that would be like. But no matter where you are in the world, a customer was a customer was a customer and all of their individual layers.

in each of the countries spanned out from that. But the customer record was almost a sacred thing, which was great. So a lot of these, I'll call them the challenges that we've been talking about, right? Technological, regulatory. cultural they're not new they've all been around for a very long time and carriers have tried to solve those challenges in various ways does ai change the game in that does it does it

level set or level the playing field for any of these problems where they no longer really become a challenge because of the capabilities? Or does it just shift the problem somewhere else? It could be just removing the barriers to that data exchange, to being able to, yeah, change the workflow the way that it's historically been done to be more customer-centric. Yes. And I'm not sure what truth there is to this, but I've often thought to myself,

Is the future of policy administration no policy administration? Is the future of billing no billing? If we go back to our AI changing the world as we know it. If you're a large carrier, I believe you probably need a core admin platform, not necessarily for the technology, but from a business perspective to have a 2B state for those 1, 2, 3,000 people that work in IT to know the target.

at where it's going. If you're a mid-tier or adventurous carrier or MGA or MGU that's got a strong technological background. You might sit there and go, we can build this with a data platform, a workflow engine, and an engagement engine with links out to Indico, et cetera, et cetera, et cetera, to bring things together. And I'm starting to see that adventure.

play out a little bit. If you remember a few months back, probably this time last year actually, Satya Nadelli came out and said, a generative AI would destroy SaaS the way we know it. It was a huge comment. And I think there's lots of truth and ideas around that. I've heard loads of podcasts of that since. I'm fascinated by it, but I still stick to the, we kind of do need that to be straight and rigor.

I don't think everyone's going to go and build their own from scratch. I think we've all advised against that over the years. We've all done that over the years. But have I got enough of the capability to give me the flexibility I want without getting locked into... another 10, 20 year project or lifespan for a platform. So I think it's really interesting. And equally, it might not be an all conversation, it might be an and. And Tom, that goes back to your point about

governance and lines of business and things geographically, it might not be one size fits all. And maybe the new AI world allows us to do, hey, for our fast flow scale market, we're doing X, but for some of these... tier two or developing markets, we're going to do something else that's a bit more agile and cost effective. I actually remember back in my Deloitte days, someone asked me for a implementation effort or ideas for a policy administration system. And they gave me their

their things they wanted to write. And I said, the answer is probably Microsoft Excel at this stage. You don't need much more. So don't overthink it early on. Yeah, the build versus buy thing is interesting, right? Because I think the idea of...

That's sort of the end of SaaS, you know, that Sascha mentioned. But build versus buy is dramatically different now. You know, I think 10, 20 years ago, you could actually hire Deloitte and build. software from scratch and basically own 100 of the code and and control your destiny that obviously brought with it all sorts of tail requirements around maintenance and and having to keep it current and so forth but

The reality is today, there is no such thing really as build versus why, in my observation, because no matter what, you're tethered to something, right? Even if you write your own code, you're going to be basing it on somebody's LLM. And so you still have some kind of vendor dependency or locking, if you want to call it. Although perhaps we get to a point where, you know, there's four or five legitimate LLMs you could run.

Maybe you try to go open source. But how do you think about, because obviously ServiceNow is a SaaS provider, where's the build versus buy discussion? I still see it very frequently. I don't think it's gone away by any stretch. the context of it seems to have shifted pretty dramatically i agree with that we saw that when i said google we see it now our whole perspective of the world is around optionality so we talk about

Any data, so we have this whole data fabric capability and data networks that we can connect to. So built-in connectors to Snowflake, Bricks, Mongo, and there's loads of others that you might get. Any cloud, so we've got the optionality of using... one or multiple clouds.

any industry, as you heard me say at the outset. So for me, I want to be that thread that goes through end to end, but at least give you the optionality. So back to your point about governance and countries, you might say, hey, I'm going to use Gemini in North America, but I want to use... open AI and Azure in the UK or European countries. And I think people want that optionality. Maybe build versus buy is not about which one's right or not. It's about not being locked into one forever.

So I don't think that debate has gone away. I remember very early on in the Gen AI, we just released Bard and then obviously that came into Gemini. People were going, we're going to build our own LLM. And I think the reality of that is when you see the costs of... I think the most ridiculous comments at the time were, oh, we'll just retune it or we'll do this or that.

The reality is when you're doing that at a hyperscale level, whether it's Amazon, Microsoft, or Google, is that a really expensive, complicated task that you wouldn't want to do. What I do think you'll end up with is a... frontier model of sorts, but then something that's unique and special to you within SLM that's relevant for your business, your line of business, your country, and lots of little ones of those will actually end up.

having more of an impact for you combined with the LLM in that space? So is the answer to that, Tom, a bit of both? I think so. I think there is no longer... two poles, build or buy. I think it's a continuum. I agree with the kind of point you made that when I talk to prospects about build strategies and ask them what drives that, there's a vendor lock-in concern.

but you're still locked into some vendor. I think that's why you see the big players will have, to your point, multiple relationships with LLMs. No one's going to bet on just one. That's not going to happen. And they don't need to, right? I think the switching costs there are very, very low. We'll talk in a minute about how you govern and measure and make sure they're actually functioning the way you think they are. So I think that's true. I think also maybe we're in a little bit of a...

The same way we had the sort of hot minute where people were going to build their own LLMs, that ended quickly. We're also right now, I see in these customers, a default desire to build first, right? Let's build everything. Let's build CRM. Let's build, you know.

build everything and i think that's going to come back to more more of a neutral position as to well where can we really gain competitive advantage and i think that i'll pull on the thread that he just mentioned which is i think the competitive advantage becomes the fabric that you put in place to

to snap all these things you know and have you create your own system that matches your business requirements or even some business requirements you mentioned you know geographies and and and whether it's you know S&D commercial or large commercial, etc. But that fabric, I think, becomes a competitive differentiator to build. But building the components within it, I'm skeptical. That doesn't seem...

a way to really establish competitive advantage. It would be so short-lived, right? Because the market wouldn't take that. I think the future is orchestrated, not built, in the nicest possible way. Like, how do you orchestrate things in a really fast, agile way? that still gives you that optionality to switch things in and out as you see fit or for different products or for different markets. I used to use a slide a while back saying, what will come first?

an insurer that makes a car or an auto manufacturer that makes insurance. And it felt like we always had that debate about embedded insurance, if you remember a few years back.

I think it's going to be table stakes going forward. It's not for everything or everyone, but it's easy for us to create a... us the industry to create a headless product that's then embedded into your travel experience your auto experience or whatever it might be there's no need for insurers to go hey let's go make a car when they've never done it before before in the past that's definitely where we should be

partnering or orchestrating the right components together. I'm curious, Nigel, because you mentioned it.

earlier on, just the speed of deployment on some of this. And you called out MGAs or, you know, either more tech advanced or nimbler, faster organizations. In my day-to-day, I've seen a lot of these AI-driven platforms or capabilities where you are piecing together these workflows outside of a... core platform. And those go-to markets are primarily focused on MGAs or very small.

carriers that are maybe monoline that are very tiny versus the big carriers. And I wonder if that's a play to just say we can... build market share that way and then prove that we can do it? Or is there just this concern among the bigger carriers of moving too fast on that or deploying too quickly across some of that and being too...

too one-off and adding too many different solutions. I don't think it's a concern that they're moving too quickly. Quite the contrary. I think every big carrier I engage with wants to move faster, do things quicker, better, faster, cheaper, all the things that go with that tagline.

I think the beauty of MGAs, specifically some of the tech-enabled ones, and there's lots of roll-ups right now as well going on, is they do have that agility and they do have that speed and they can go capture market. I think the M&A space right now... And the private equity space seems to be really interesting and acquisitive to bring lots of these folks together. What will be interesting is when you've bought, you know, not just...

10 together, but 100 together, whereas the, do you eventually end up at a gargantuan scale that means the speedboat all of a sudden... morphs into a tug morphed into an oil tanker and we've all used the analogies over the years i suspect lots of the large carriers and institutions whether they're carriers or brokers to be fair have a

oil tanker and speedboat strategy that says, for our main business, it's going to be like this. And we appreciate that it's going to take time to implement change and elsewhere. But for some of the extremities, we can do things that are ands, not oars. And I think that's probably the new reality. And if I go back to our conversation on data, if you've got lots of ands,

then the data becomes even more important again because you can still have it all pushed back down to Tom is Tom no matter what piece of business he's doing. So you might have different interfaces in the middle to do that, but that's where the data continues in my mind to be.

Absolutely imperative to speed here. What do you think in an agentic world isn't being talked about enough that it's going to surprise the enterprise as it tries to scale and roll out? an agentic world. What do you think is not getting the attention of these? We touched on it briefly earlier. I've got two things that are top of mind in almost every conversation. One is the... the controls and the governance around what's going on, because I don't think one model will rule them all a one.

custom GPT will rule them all. I think there's going to be hundreds, if not thousands, across the organization in the same way there was applications. So getting to grips with that and putting control mechanisms in place. to give you the confidence, specifically in regulated environments, is critical. It's going to be table stakes. We're doing some really cool and interesting stuff in that space around something we call AI control tower. I'm sure others will follow, but that to me is imperative.

specifically as we're trying to get ahead of it, rather than play catch up like we've done in the insurance industry for the last 20, 30, 40 years. That's number one. The other one, and it is talked about, but it comes back to enablement. I think the push from top down on this initiative...

is phenomenal. Every CEO, every board is talking about AI transformation and taking costs out and being more efficient. I think probably one of the most important people or roles in every organization right now is the chief. HR officer or chief people officer, not from what's going on with jobs displacement, of which there will be some, but equally the job augmentation. What does my new role look like in an AI-empowered world?

What does enablement look like for my staff that have been on board one year versus 20 years? How do I encourage my net new people coming in, like the kids are 12 and 16, and my 16-year-old uses ChatGPT and Gemini and perplexity? all the time. It's a whole new language and era. That's what they've grown up with. In the same way that we talked about the green screen earlier, you go back 40 years and we were all used to green screens and we were bringing in new folks who were used to using...

physical apps, then web browsers. And it was a change. I think we go through that change again. And there's some really great examples out there. Canva, the Australian company that does the designs and PowerPoint equivalent, just gave everyone an AI week. Go explore.

Go understand what the tools are and how they impact your job and what you do. We've seen the same with folks doing coding with Cursor. ServiceNow has got regular AI days for everyone, for every one of the 27,000 people. Today's an AI day. Go explore. or go be curious, go play, see how it works, what doesn't work, and how it applies to your job. And then folks like, in fact, even here, Jackie, our chief people officer, is now our chief.

people and AI-enabled person. She's responsible for making sure that we're leading from the front. You might say that's easy for a technology firm or the big tech space out of California. My other favorite example is Moderna. I think Moderna just merged. their HR and IT functions, specifically with a focus on making sure that the tools that we're providing are in use and enabled.

throughout their entire organization. So I think you're starting to see a merger or an interesting merger. I've not seen enough to tell you it's a trend, but I'm definitely intrigued by making sure people are ready for this. Yeah, it's fascinating.

This has been a terrific conversation, Michelle. We've been talking to Nigel Walsh, the global head of insurance for ServiceNow. You've been listening to another episode of Unstructured Unlocked. I'm your co-host, Tom Wild. And I'm your co-host, Michelle Govea. Thanks, Nigel, for joining us. Thank you very much, folks. Thank you for joining us for this episode of Unstructured in Love. You can find all of our episodes wherever you listen to podcasts today. Spotify, Apple Podcasts, anywhere.

Be sure to write a review if you like what you hear.