SortSpoke Blog » Latest Articles
Feb 16, 2024 SortSpoke

Live at ITC: What is TriageAI?

SortSpoke Founder, Jasper Li addresses the talent crisis in Underwriting today, Augmentation vs Automation, and SortSpokes’ Triage Engine.

 

 

Moments to listen for:

[00:00] Introduction

[00:40] Use Case: The Talent Crisis in Underwriting

[01:42] Why SortSpoke Was Founded

[02:15] SortSpoke Results

[02:26] Augmentation vs Automation

[02:47] How it Works- The Problem

[03:49] How it Works- The Solution

[04:58] How it Works- An Example

[06:38] How it Works- Triage

[07:11] Who SortSpoke Serves

 

Video transcript:

[00:00] Introduction

Good morning everyone. SortSpoke is an AI-powered document processing platform purpose-built for insurance. We help underwriting teams intake, triage, and extract data from all the submissions they need for underwriting. We mostly serve P&C, Commercial, Specialty, Life Carriers, TPA's and MGA's. I'm Jasper Li, the CEO and Founder of SortSpoke. Prior to founding the company, I spent about 20 years working with carriers and financial institutions to improve their operational processes. And I wanted to share a little bit about the story of SortSpoke and how it came to be. 

[00:40] Use Case: The Talent Crisis in Underwriting

So this is Sally. She's a senior underwriter. She's maybe one of the best underwriters on your team. Maybe even in the country. She knows how to look at risk, she knows how to find, and how to price. But she still spends about 30-40% of her day just reviewing documents, pulling data out, and doing clerical work. But you had 100 Sally's you could take over the market. And by the way, Sally's been doing this for about 30 years. She's probably pretty close to retirement.

This is John, the new assistant underwriter. You just recruited him from a top school, super smart, top of his class, and super eager to learn how to be the next generation of underwriters. You know he joined you because you're a great company and you invest in your people and your technology. He probably spends 70% of his day doing intake and data entry, just to save Sally a little bit of time. That's not what he signed up for either when he joined your company. This is the talent crisis that we're seeing in underwriting today and why underwriting, at least for more complex commercial life products, is ultimately constrained by your underwriting capacity.

[01:42] Why SortSpoke Was Founded

That's why we found it SortSpoke. Because what if you could give them an AI that helps them take some of the drudgery away so that they could focus on the parts that the humans really should focus on? Which is what if you could give them SortSpoke? Well, you could win more new business, ultimately by increasing the capacity of those underwriting teams, using the AI where it makes sense to extract the data from those documents, prioritize which submissions are the most on appetite risk, and then present that back to your underwriting teams so that they can do the work that they should focus on.

[02:15] SortSpoke Results

Typically what we see is about a 50 to 70% increase in gross written premium per underwriter. Just a raw capacity increase of about 30% in terms of submission capacity.

[02:26] Augmentation vs Automation

And this is the important part, that we founded philosophically the entire company on- we're not trying to automate your people away. You want to make them happier. You want to retain them longer and onboard your next generation of underwriters more easily. So we are all about keeping your people happy and keeping them in the loop. We’re not here to try to get rid of people, we don’t think that’s the way forward.

[02:47] How it Works- The Problem

This is probably what a lot of companies look like in underwriting teams; in this case, they receive 16 submissions a day. The thing I want you to focus on is there are 7 bubbles over there. The 7 bubbles are the submissions that if you could quote those submissions within 24 hours you will win them because you have the best product and the best pricing. The problem is do you have manual intake data entry and logging into your system, you’re probably losing about a third of your submissions because you can’t even get to them by the time new submissions are coming in tomorrow. Of the 7 you could have won, two you lost cause you didn’t even open the email and get back to the broker. Of the ones you do actually get the quote to, you probably lost two more because it took you a week or more to respond, and by then the insurer has already moved on or the broker has already pushed them to someone else. This is what we kept hearing over the last 20 years and why we developed SortSpoke. You’re losing the winnable submissions just because you’re not getting to them in time.

[03:49] How it Works- The Solution

With SortSpoke, those same 16 submissions come in, there are the 7 winnable ones, and within seconds of the submission coming in the system will pull out enough data to be able to triage and say this is the highest quality risk, these are the ones you should focus on. Immediately within a minute, these will get sent to your underwriters to do a quick QA, the system might say here are the 80 data points that you need to underwrite and there are 2 I really need you to check because they filled it out in a funny way. There are always exceptions, submissions are messy. There’s always a human in the loop, we leave your underwriters in control. Same day you respond and you quote- guess what you won the 7 winnable ones because you got to it in time. And you have so much extra capacity that maybe you even respond to a few extra lower or medium-priority submissions and maybe you win some of those too. And then a third of off-target risk, where the broker forgot to submit the loss run, are just automatically sent back to the brokers, you don’t need to waste your underwriting team's time on them. That’s what SortSpoke does, and that’s how we can increase 50-70% of gross written premium per underwriter.

[04:58] How it Works- An Example

Let me show you what that looks like. So intake, whether it's through emails or an API directly through the portal or whatnot, in this case, an email with attachments and half the useful information is just embodied in the email to the broker rapidly typed out. You send that to SortSpoke, we extract enough information to do initial triage and then we sort them into the different buckets (these buckets are up to you and your risk appetite) but the first thing is these are all the high-priority submissions that came in today and you as an underwriter will claim the ones that you want to do. SortSpoke will say here’s all the data you need from the submission already extracted but it’s going to guide the underwriter to do some degree of human review because there are always exceptions. It’s going to say, the description of operations was embedded in the email, I need you to look that over, you shouldn’t trust the AI here to do it correctly. It’s gonna pull you to other supplemental documents in there as well, and just say this is where I need you as a human to check the AI, because you should never fully trust an AI. We make that really easy to do and all of the predictions that the AI does are fully explainable, there’s no trusting figuring out how it came up with it, we show you exactly where and exactly what you need to do to make sure that you have all the data now ready for submission. And that’s it- we would push that directly into your underwriting system, your rating system system, your policy admin system at a point. 

[06:38] How it Works- Triage

On the Triage side, we have a rules engine built in. It’s not a replacement for maybe your existing rules engine or rating system, it’s really just to set up some rules. So you could say if you’re in food service can you operate anywhere but New York State and you have between $10-$50 million in revenue that’s a Green, I want that underwritten today. That’s the kind of thing that you would do with SortSpokes’ Triage engine. And all of that’s explainable it’s not a magical AI blackbox.

 

[07:11] Who SortSpoke Serves

SortSpoke mostly serves large Commercial P&C, Specialty (especially where there are lots of supplementals and complex products and bundled submissions, as well as a life worker’s comp, and things of that nature). Think of us as a people to go to when you have complex unstructured submissions, and you have people manually doing it today. We do the hardest submissions- that’s our niche. We help you prioritize them, and we get up and running very quickly (within weeks you can be a production). Usually, the bottleneck is just how easy is it to get the data into your policy admin or under any workbench, but if they’re relatively open very very easy to do. I want to emphasize, that our philosophy is you’re never gonna achieve straight-through processing with these complex submissions. If you need 100% data quality, you have to have your humans in control. SortSpokes AI makes that explainable, so for every prediction the AI makes you always know why it made it, and where it came from so you can override it as soon as you override anything the system does it learns from it in real time and it knows how to deal with that situation next time.