Get a demo
Select the type of demo you'd like to experience with Indico Data.
Select the type of demo you'd like to experience with Indico Data.
Introducing season 2 of Unstructured Unlocked! Indico Data CEO, Tom Wilde, steps in as co-host alongside Michelle Gouveia, VP at Sandbox Insurtech Ventures. In this episode, they discuss Indico’s recent top position in Everest Group’s Intelligent Document Processing (IDP) Insurance PEAK Matrix® 2024.
Michelle Gouveia:
Hi everyone. Welcome to a new episode of Unstructured Unlocked, Michelle Govea. I very excited to be welcome today by our guest and future co-host Tom Wilde, the CEO of Indico. Hey Tom.
Tom Wilde:
Hey Michelle. How are you?
MG:
Good, how are you?
TW:
Great.
MG:
Very excited to kick this off with you. Wanted to just take a moment though and let the audience know. We, as I just mentioned, will be transitioning to having Tom and I as a new co-host of the Instructor Unlocked podcast. We really want to thank Chris Wells for all of his contributions and for really kick-starting this podcast and for getting it to where it is today. So major shout out to Chris on that.
TW:
Yeah, agree. Super content that you both produced over the last year. The podcast has grown so much, so really excited to be joining you and to continue the legacy here of producing interesting content around all things unstructured data.
MG:
Phenomenal. Great. So without further ado, I’ll jump into the topic of the day, which some of our more seasoned listeners may recall being a similar topic from a former episode. But it’s that time again where Everest has released its peak matrix reports. And so wanted to dive into that again with you today and just talk about what goes into that report, why is it important to be on there and just talk a little bit about the differences and new things that are in there from last year to this year. So I think maybe to get it started, Tom, do you want to talk a little bit about what is the Everest Group IDP report and from Indico standpoint, why is it so meaningful to be included in it?
TW:
Yeah, that’s a great question. I think for those who follow IDP and I kind of think of the folks in the know around IDP, they’ll have recognized that Everest was very early to really carefully dimensioning this space and reporting on the vendors and the trends and they’ve continued to do so. I think we’ve been working with Everest in some fashion for the last four or five years at least as a participant in these reports. And I think it’s seen as the gold standard for laying out the IDP landscape.
MG:
There’s some folks out there that may think that these are all just pay to play and the list of companies that’s in there, they bribe their way in their a better description. What does it really take though to get into the reports? I know that that’s not the case and I know that it takes a lot of hard work both from obviously the development of the product and the company, but also just building relationships. So would love to pick your brain on that too.
TW:
And I’ve been working with analysts for 25 years now in the technology space. So I’ve got a lot of perspective and experience on how the analysts approach these markets. Specifically when talking about vendors in the market. Primarily you have to remember the analysts main customers are the end users. They’re the big enterprises who are trying to get smart about where the emergent technologies are, where the emergent vendors are, key strategies for adoption success, all of those kinds of things. Now, the analysts also have relationships with vendors from time to time in terms of the vendors can subscribe to the analyst research and have analyst inquiries to make the vendors smarter about the markets they’re targeting as well. My experience for the last 25 years is there’s a very strong firewall between those two things. No analyst report on the sort of the way we’re describing it that I’ve been a part of has ever required any kind of payment to be part of it. There is custom research, which I think sometimes gets confused. There is custom research that some analysts will do where you do pay them to write a particular custom research on a particular technology or market, but that’s always disclosed when those reports come out. So custom research is different than these analyst industry reports.
MG:
One of the things that I always find impressive in these reports is the mix of the type of companies that get in there. There are some earlier stage companies that are in the same kind of boxes or categories as larger incumbents or companies that are household names, so to speak, in seeing the companies that were in the report, and some were there last year, I think there are some that are net new. How do you think about it when Indigo was in the mix with both sets with companies from both sides of the spectrum?
TW:
Yeah, I mean the have to balance a couple things. They’re trying to make sure that the vendors they present to their end customers who are going to read these reports that they’re picking up on emerging companies, fast growers and established vendors. So they want to cover that spectrum. They want to be careful not to cover brand new startups because they don’t have enough evidence of traction and success yet. So they’re a little cautious on that, but they do want to make sure they’re covering emergent players who seem to be gaining traction. I think the other thing is that you’ll see some companies come and go from these reports because they reposition what they offer in the market. One thing that has been sort of remarkable about Indico, and I say remarkable because I’ve built many startups, it’s really difficult to maintain or to create a positioning that you can maintain year over year.
TW:
Generally, you have to do some significant pivots when you’re a startup. What’s interesting is Indico has been positioned as an enterprise AI solution for unstructured data from the very beginning. In fact, I found an old mouse pad in a box that I was unpacking from 2018 and it said enterprise AI for unstructured data. So that’s how long we’ve been focused on this space. So part of the movement on the reports is companies coming and going in terms of their relative focus. In this case, IDP companies get acquired or go out of business. So that’s part of the reason why you see movement in terms of who gets plotted, where in this case the peak matrix.
MG:
And I mean from a VC standpoint, these reports are always helpful too to just keep a pulse on how is the landscape shifting, who are the new players coming in, who are potential acquirers for companies that we have invested in this space? So there, I’ll plug that for the VC angle too, that it’s super helpful. That’s true. Yeah. The one, not difference maybe, but deeper dive this year is that the specification of insurance in the report, it’s an insurance specific report that they put out, which was not the case last year. I think Indico was in the leader column last year and is a broad category and now it’s in that same leader box. Again, congratulations. I’ll just keep popping that in there, but insurance specific angle. So why is that meaningful, especially given how you think about Indigo’s positioning and how the company has moved really to putting insurance use cases in the value proposition at the forefront of the product development and everything like that.
TW:
And I’ll even go back in time a little bit to talk about the arc of this report going back, call it four years. So the first time we participated in this report, we had a lot of deep discussions with Everest about how they were defining intelligent document processing. We felt that they were not paying attention to the emergent unstructured end of the spectrum that the original report focused a lot on form-based, we like to call it structured and semi-structured documents, forms, invoices, things like that. And they weren’t focused enough on the emergent unstructured space, which was really at the time and still to this day, powered by deep learning foundation models, which we now all call large language models. So that was Indigo’s forming legacy was that we were the first in the market to use a foundation model and discrimative AI to solve this problem.
TW:
And now of course, generative AI as well. What you saw over the next several years to Everest credit is they began to subsegment the IDP capabilities into unstructured semi-structured, structured. And that’s accurate. That’s the right way to think about the various capabilities of the vendors in the space because that’s how the customers experience the problem. They have those sort of three categories of documents that they’re trying to solve with some kind of document understanding technology. So it makes sense that this year they took the next logical step, which is to add some reports that are more vertical specific. I think they did banking and insurance this year as additional reports. The reason for that is you find that the use cases in the industries can be quite unique. And so the domain expertise of the vendors as it relates to those industry use cases is very relevant for the readers of these reports, which are namely the big enterprise customers in insurance, in banking, in healthcare, et cetera, who are trying to figure out, okay, IDP is sort of a horizontal technology, but will it solve my specific industry problems?
MG:
I want to go a little bit deeper on this, and maybe this question’s maybe a little less about the Everest report itself, but because you said the magic words in that answer, right? Large language models and generative ai, there is so much demand, especially in the insurance industry to execute something within the four walls of the insurer too, to say, we have successfully leveraged generative AI or we’ve automated these processes using ai. There’s that need to explore it and be successful with it. How does indico sift through and navigate through all what I’ll call the hype and maybe the unproven solutions and really just stand out, as you said, you were at the forefront of using ai, you’ve been using it at this core to the solution for as long as Inco has been around. So what does that process look like to really identify what is the outcome you’re trying to solve and how is inco responsive to that?
TW:
And I think this is where the analysts really have to focus every year. There could be some significant technology innovation and they have to map that onto, in this case IDP, and understand from each vendor how they’re thinking about that. And so I think that’s one thing that I’ve found to be really fulfilling is that we’ve stayed, we’ve grown into that leader quadrant and we’ve stayed there. And that’s because we have to innovate as fast as the market is, if not faster, because the analysts really interrogate the vendors as to how they are adapting to new technologies and adopting and deploying those to their customers. So the IDP landscape has changed with the arrival of Gen ai. I kind of break the IDP problem into three parts. So you have extractive problems, you have interpretive problems, and you have predictive problems. Now with Indigo’s innovation, using a foundation model, we really set the bar around extractive.
TW:
So that’s classification, extraction, those kinds of capabilities that allow you to turn unstructured data into schematic structured data. Predictive is probably the oldest part of the landscape, which is traditional machine learning. So that is statistical regressions described most simply, and that’s been around for a long time since the sixties. The middle piece is what’s arrived here with the dawn of GPT, which is interpretive, what large language models allow you to do. And in the case of these hyperscaler large language models, we’re talking about trillion parameter large language models, is now for the first time we can actually do interpretive type use cases using NLP that is now supercharged by these LLMs. By interpretive, I mean things like summarization, supplanting what we used to do with rules to a certain extent, those are all interpretive, right? Creating dynamic table of contents from very long documents or from a set of documents. These are all interpretive use cases. So it’s a really exciting time in the IDP landscape to have those three now very much in focus and the vendors who are leading the market are able to do all three. And we certainly feel that we’re one of the leaders in this space with capabilities across all three.
MG:
For sure, for sure. And I think in reading this report, maybe it is because it went so deep on the insurance vertical, or maybe it’s just a new format, but they had a few pages that went much deeper into the capabilities of indico and I’m sure all the other companies as well. For those specific readouts, what do they have? What’s on the roadmap and how strong is that offering? What is that process like to have them vet all of those core capabilities as part of this report in readout?
TW:
Yeah, there’s sort of like the vertical and the horizontal have to come together there. From a vertical perspective, what all customers want is to know that you’re very, very familiar with their business problem. Forget the technology for a second. Do you as a vendor really have experience with the business problem I face? And do you have a track record of solving that? Right? So that’s where a deep understanding, let’s use insurance because it’s the specific example here of things like very specific document understanding challenges around underwriting, around claims, around policy servicing, around financial operations, where these are all very unstructured centric, document centric business problems, and the customer has a specific outcome they’re trying to achieve. If it’s underwriting, can I more quickly get at the right kind of risks that I want to write and can I quote those as quickly as possible that allows me to win more business, that allows me to grow revenue with claims.
TW:
Can I make sure I am adjudicating the right claims and doing it quickly? Am I making sure I’m avoiding fraud? Am I servicing the customer and delivering on the promise of the policy that I made with them so that they stay with me as a customer? And can I increase the efficiency of how quickly and how much it costs me to process those claims on policy servicing? Very similar. My customer satisfaction depends on me being able to respond to a policy holders updates and changes. These things change constantly. If you’re a big marine shipping line, well, they buy and sell ships and change ports of entry and all these kinds of things, and that has to be continuously scheduled and updated. So how are you able to do that accurately and robustly to maintain customer satisfaction as well as make sure that your policies are up to date.
TW:
So customers would want to know that you have an understanding of those things now that understanding extends into the technology because they also want to know, do you have any kind of accelerators or out of the box workflows or starter packs that allow me to get to value quicker? Because the time to value is a real challenge in the IDP space, these are complex use cases, often involving now artificial intelligence, which means there’s model training and workflow development and integrations with upstream and downstream systems. So they are also wanting to understand that you can get them to value quickly because they’ve made commitments to their internal stakeholders.
MG:
And I want to call back to a part of the conversation that we had in our previous episode when we were talking about the ever adverse report, where you made a comment about as part of sometimes the sales process and going in with Indico, that the prospective customer usually has the pain point and then they have in their head what the solution would be. And it’s essentially like, can you do this for us? And you made a comment that sometimes it’s really us going in and saying, well, hold on a minute. The real outcome that you’re trying to achieve is X. And so if we went about it this way, it would actually get you closer to meeting that need. How do you think about that from all the use cases that you just talked about and now generative AI where there’s a lot of times where people are saying, well, I can just use AI for that. What’s the back and forth a little bit of let’s solve your actual problem versus just throwing AI at the problem? And what is the dynamics there when you’re talking to some of the folks in that room?
TW:
Yeah, I think there’s a reason it’s called artificial intelligence, right? It is artificial. What I mean by that is while the sort of concept of a digital twin makes sense conceptually that, Hey, here’s how the process works today with the people that are doing it. If it’s a very manual existing process, it’s not quite a direct lift and shift onto ai because AI is artificial. You have to kind of break the problem down in a way that is consumable and solvable by ai. So there’s some experience, skill and thought that has to go into that to make it successful. There’s a reason that 40% of AI projects fail to get to production because those steps have to be really considered because it’s not just, do I have a model that can extract these 10 fields from this document that’s relatively today relatively easy to do?
TW:
It’s all those things around it, which is, how will I do human in the loop? How will I be able to explain the models and their performance? Can I create an audit trail around every prediction that was made, every extraction that was made? What is my throughput so that my turnaround time on these documents processing meets my business requirements? What is the impact of an error? How do I remediate an error? What systems do I have to connect with upstream and downstream? So it’s a fairly sophisticated solution when you unpack the whole thing. And so really starting with a conversation with a customer, which is what outcome is going to deliver value, when your boss’s boss asks, why did we do this? What are you going to say as the reason you did it? And then work backwards into how to solve it, it’s understandable that people want to start with, here’s how we’re going to solve it without really knowing or having enough experience in how to weave AI into these problems. And that’s not just in IDP, by the way. I think if you look at the challenges you’re seeing in some of the funnier stories around chatbots right now with gen AI and some of the things gone wrong, chatbots gone rogue, similar kind of problem. It’s not a direct lift and shift from the way it’s been done manually to a way that an AI will try to do it.
MG:
And I mean, you know this add in all the compliance and regulatory requirements on top of all of those kind of internal processes and checks and balances. And like you said, it just becomes this major, not complicated in a sense that you don’t want to do it, but it, it’s a complex execution to get something fully implemented. There’s a lot of moving parts.
TW:
The Air Canada story was one of the most peculiar where the Air Canada chatbot, I think sold the customer a flight at the wrong price and quoted incorrectly, air Canada’s policies around fees incurred for changing that flight. And the customer took them to court. And Eric Canada’s argument was, well, that was the chat bot that did that. And of course the court said, well, that’s ridiculous. It’s your chat bot. So we’re just at the very beginning of things like that, surfacing and enterprises realizing that the testing, the deployment, the explainability and the auditability of these things is vital.
MG:
Yeah, well hopefully, well, if I’m looking on an airline chat bot, hopefully they give me something for free. Going back to the report, because there are so many, this landscape is moving so quickly and there’s so many players, what do you think looking out to the next pick a timeframe, 18 months, two years, is that next big push? Where do you think that the needs or the capabilities of technology will be executed in relation to the insurance industry specifically?
TW:
I think all customers generally are wanting to, and I’ve had customers say this to me directly, they’re hoping that the vendor solutions can help them to simply get to the decision at the end, just jump over all this, just give me the right decision for my business process. Should I approve this claim? Should I underwrite this risk? So that’s the answer they’re looking for. So they’re, I think, increasingly really trying to understand how you’ll get them to the decision, not necessarily just the data. I think at Indico we’re starting to aggressively talk about the beginning of the decision era. Nobody needs to be told they need a data strategy that’s a 20-year-old thought at this point. So if I said that to somebody, they would look at me like I was out of my mind, and if they didn’t have a data strategy, I would look at them like they were out of their mind.
TW:
So it doesn’t matter what industry you’re in, everyone is a data company, no more so than insurance companies. I mean, if you think about insurance companies, IP is really their decision making capability. That’s what it is. They have a big balance sheet and they’re really smart at making certain kinds of decisions. That’s an insurance company making promises based on those decisions to their customers. So increasingly, the customer’s hope that because of the arrival of some of these new flavors of artificial intelligence like gen ai, they’re going to be able to get to that decision in a more automated, more robust fashion. We’re not there yet, but they’re thinking about it the right way. And as a vendor, we have to be thinking about how we’re going to do that for them, with them as sort of the next wave of these technologies.
MG:
For sure. And I’ll kind of wrap it here with this final question. What do you want people who read the report to take away from it about Indico or about the industry or about where Gen AI is leading in both of those things?
TW:
I think the thing that we’re most proud of, I’ll kind of repeat what I said earlier, the durability of ICO’s portrayal as a leader on these reports over the years. If you’ve been on the vendor side of the equation, you know how hard it is every day. Someone’s trying to eat your lunch. There’s new startups every day. There’s big giant public companies who will decide to enter the space. It’s very difficult to maintain a leadership position on these reports. The analysts are vigorous, they’re paid to be skeptical. You have to demonstrate your qualities as a leader. So I think we’re very proud of that and I think it’s a real data point for prospects and customers to pay attention to for sure. I think that in addition, the arrival of this vertical specific reporter and insurance indico has focused a lot on insurance over certainly the last two to three years and have become expert in the use cases that our customers are challenged with in insurance. And I think that we continue to invest heavily in really delivering unstructured automation and analytics for insurance and financial services.
MG:
Excellent. Well, Tom, thank you so much for taking the time to chat with me today. This has been another episode of Unstructured Unlocked Michelle Govea, and I was joined today by the CEO and Future Co-host of Unstructured Unlocked, Tom Wild.
TW:
Looking forward to it. Thank you for the conversation. Looking forward to the future here.
MG:
Excellent, thanks.
TW:
Thanks, Michelle.
Check out the full Unstructured Unlocked podcast on your favorite platform, including:
