Conversations on the Creek · 2024-05-14 · 10 min
Key moments - from our scoring
Substance score
27 / 100
Five dimensions, 20 points each
Andrew Schwartz of Celent identifies claims management as a critical area where generative AI can drive immediate value for insurance carriers facing persistent loss ratio challenges and increasing natural catastrophes. The lowest-hanging fruit for AI adoption lies in automating first notice of loss processing - where generative AI synthesizes structured and unstructured data to give adjusters a complete picture - and in intelligent claims triage to distinguish low-risk from high-risk claims requiring human expertise. However, successful implementation requires three foundational elements: change management to ensure adjuster buy-in and comfort with the technology, mitigation of AI hallucinations through smaller, well-curated training datasets rather than massive general-purpose models, and adherence to the NAIC model bulletin on AI, which has already been ratified by six state insurance departments. Schwartz emphasizes that human-in-the-loop review remains essential, and that carriers must ensure third-party vendors meet the same governance and risk control standards they maintain internally to avoid regulatory exposure.
First notice of loss processing and claims triage are the lowest-hanging fruit - generative AI can aggregate all documents and data to help adjusters make informed decisions and route high-risk claims appropriately while filtering routine claims.
Use smaller, pre-trained datasets tailored to your specific claims domain rather than large general-purpose models; this reduces false information and gives carriers more control to understand the impact of changes.
Carriers are now responsible for the actions and governance of third-party AI vendors they work with, requiring careful vendor selection to ensure they meet the same risk controls and compliance standards as the carrier.
Generative AI models can produce hallucinations and false outputs, so human adjusters must validate AI outputs before processing claims to catch errors and ensure accuracy.
Invest in change management and education to help adjusters understand that AI tools enhance their work rather than threaten their jobs, and position the technology as freeing them to focus on policyholder interactions instead of data gathering.
Our reviewer’s read on each dimension, with quotes from the episode.
The episode is only 10 minutes and covers almost entirely surface-level observations - GenAI helps adjusters, hallucinations are a risk, data quality matters. The only moderately non-obvious point is about carriers using smaller pre-trained corpora to reduce hallucinations, but it is not developed with any depth.
we have seen a lot of carriers think about working with smaller data sets that are pre-trained so they can, A, understand the impact of different changes within that, and also, B, because of the fact that there aren't the same kind of hallucinations
first notice of loss, being able to pull all of the different documents in a claims situation together so the adjuster is able to make some sort of a meaningful assessment
Every idea here - human in the loop, change management resistance, hallucinations, regulatory uncertainty - is a stock 2023-2024 AI conference talking point. Nothing is contrarian, first-principles, or counterintuitive; the guest explicitly echoes the 'phrase of the conference' the host identifies.
I think, phrase of the conference, is human in the loop
not necessarily to take their jobs, but it to help them actually do their jobs in a way that is better
Andrew Schwartz is an analyst at a research and advisory firm, giving him breadth of industry observation but no direct practitioner experience deploying GenAI at scale inside a carrier. His commentary reflects conference synthesis rather than hard-won operational knowledge.
Sellent is a research and advisory firm dedicated to helping financial institutions formulate comprehensive business and technology strategies
It was a fantastic panel on claims productivity through analytics. Definitely a very receptive audience and a lot of interesting takeaways
The only concrete data point in the entire episode is that the NAIC model bulletin has been ratified by six state departments of insurance. No carrier names, no metrics, no loss-ratio figures, no implementation timelines or dollar values are provided anywhere.
the NAIC model bulletin on AI which has been ratified by six departments of insurance now since it was passed at the end of last year
loss ratios have been very sticky
The host asks broad, leading questions ('lowest hanging fruit,' 'regulatory challenges') and never pushes back on a single claim. His follow-ups largely restate what the guest just said, and the conversation ends with filler affirmations rather than any productive tension.
Sure. So when it does come to generative AI, which part of the claims lifecycle do you think might be the lowest hanging fruit
Good stuff. No, I appreciate it. Transparency, traceability, visibility, and what you're looking for in your vendor
Computed from the transcript - who did the talking, and the words that came up most.
In today’s episode, recorded following a panel discussion at the Insurtech NY 2024 Spring Conference, Andrew Schwartz, a Senior Analyst at Celent, joins host Rob Savitsky to explore the use of generative AI and advanced analytics in insurance claims management. They delve into various applications, from first notice of loss to claims triaging, highlighting the potential for generative AI to enhance adjusters’ efficiency and elevate the customer experience. Listen in for an analyst’s perspective on these topics, and gain insights into the latest regulatory developments, including expectations outlined in the NAIC Model Bulletin regarding AI utilization by insurance companies. Learn more about Celent by visiting celent.com and
Transcribed and scored by The B2B Podcast Index.
Welcome back to another episode of Conversations on the Creek, the dog creek podcast where we interview thought leaders about how the latest insure tech is transforming the PNC insurance industry. Whether you work in underwriting, sales and marketing, claims or an insurer's IT department, in each episode we uncover the insights you need to reimagine the future of insurance. I'm Rob Stavitsky and in today's episode I'm so excited to welcome Andrew Schwartz of Sellant to share his perspective on applying gerundive AI and advanced analytics to claims management.
In case you have not heard of Sellent, Sellent is a research and advisory firm dedicated to helping financial institutions formulate comprehensive business and technology strategies. So without further ado, here is our second micro podcast episode recorded live from the InsurTech New York 2024 Spring Conference. Rob Stavitsky here with Andrew Schwartz from Sellant just finished a panel. How is it going today, Andrew?
It's going great. It was a fantastic panel on claims productivity through analytics. Definitely a very receptive audience and a lot of interesting takeaways. Yeah.
So speaking of takeaways, obviously there's a lot to talk about on this panel. You had a good mix of solution providers, carriers on the panel. What comes to mind when you're thinking takeaways for claims analytics? Well, I think that everyone acknowledges that we are at an important inflection point within the insurance industry.
Certainly within the last couple of years, loss ratios have been very sticky. Different carriers have tried to employ a variety of different strategies in order to combat that, and that combined with the fact that we've had an increasing number of natural catastrophes, certain number of carriers, high-profile ones, have pulled out of many different states because they just haven't felt comfortable being able to accurately assess risk. And what we've found is that at the time of all of this kind of dysfunction and kind of chaos, there are a lot of really exciting solutions out there and technologies that are enabling carriers to be able to more effectively analyze and respond to these risks.
and we've talked a lot about how certain tools, like for example, generative AI, have been really instrumental in being able to help carriers process claims in a way that is more expeditious efficient and allow them to do it in a way that also improves customer experience doing it in a more accurate way but also in a more rapid way as well Sure. So when it does come to generative AI, which part of the claims lifecycle do you think might be the lowest hanging fruit for where you might consider applying it as a carrier for jumping in?
Right. Well, I think that claims likely for the foreseeable future, for the most part, not glass claims, but most claims will be handled by an adjuster. And in many ways, generative AI will be a tool that will allow them to more effectively gather all that structured and unstructured data and actually make meaningful insights out of it. And where that can happen in the claims process, lowest hanging fruits, number one is going to be the first notice of loss, being able to pull all of the different documents in a claims situation together so the adjuster is able to make some sort of a meaningful assessment based on all the information that they've had.
And I also think triaging, being able to help an adjuster understand what claims are low risk and there may not be some sort of an issue. And also some that maybe are a little bit higher and actually would require their human expertise. Nice. Some good points.
I think people were making the point yesterday that like on the underwriting side, the importance of getting all the data together. But not only that, but the next step of like synthesizing and putting out what matters. And it sounds like some similar points with some of the document extraction and the opportunities there are to really boil down to what does the claim adjuster actually need in order to move forward and do their job more effectively. Well, I think there are a couple pieces.
The first piece is I think the technology is there in many ways, but in order to also effectively leverage the technology, it's important that carriers understand the risks that are inherent as well, which would be first the change management piece. If there's a really exciting technology and it's not a technology that an adjuster is comfortable with utilizing, then what kind of use is it going to make? And that would require educating them on that and making sure that they understand the value of these tools and they're not necessarily to take their jobs, but it to help them actually do their jobs in a way that is better and can allow them to focus on what matters and that a lot of many ways is dealing with the actual policyholder or customer So that one Another real, I think, important thing is helping people understand that sometimes these tools lie or as they're called hallucinations and understanding what are some areas that maybe they shine in and what are other areas where they might have to be a little more cognizant as well.
So understanding that there are hallucinations and also acknowledging ways that they can potentially mitigate them. So that's number two. And then I think number three as well is being able to access the right data. And at the end of the day, there's a whole comp sci principle of GEICO, garbage in, garbage out.
So making sure that you actually have a small or I would say smaller typically trained corpus of data that you're able to make some sort of inferences from is very important. And that's going to require folks at carriers to understand that they need to make sure that they have a really well understood data set that's a bit smaller, that they're able to actually feel comfortable in doing some sort of a task. And that's why we have seen a lot of carriers think about working with smaller data sets that are pre-trained so they can, A, understand the impact of different changes within that, and also, B, because of the fact that there aren't the same kind of hallucinations and there isn't the same kind of false information from it.
Sure. Good points. And I think one of the things that came up, and maybe it's the, I don't know, phrase of the conference, is human in the loop. I think that we've been at this AI conference, but it seems like in almost every case, folks are in agreement that you need to have your front line adjusters checking or whoever it is who's working your data side to, like you said, making sure that these hallucinations of the models are not running awry.
We've only got a couple minutes left in this micro episode. I know that you did start to touch on regulation towards the end there. I guess, what do you see as sort of the regulatory potential challenges that carriers should start thinking about today as they look to develop and deploy more of these Gen AI models? Yeah, well, the regulatory issue is one that is top of mind for many carriers, but it also one that is rapidly evolving It has crystallized a little bit recently because of the NAIC model bulletin on AI which has been ratified by six departments of insurance now since it was passed at the end of last year.
I think it was in December. So what that means and what we're seeing is, A, the NAIC model bulletin, which has been adopted, has said that carriers have to be very cognizant and in many ways are responsible for the actions of third-party vendors that they work with. And what that means is if you're a large carrier and you have stringent governance structures and you're a company that's been around for a long time, you want to be mindful that you're actually working with a solution provider or a vendor that adheres to the same kind of ethos and has the same kind of values and risk controls so you're not vulnerable to that.
That's A. And I think also coming down the pipeline are going to be definitely some more, I think, stringent regulations as this technology becomes more well understood. I think that that's actually a very good thing for the industry, and it's going to allow for greater proliferation of generative AI tools because of the fact that now that carriers understand what the rules of the road are, It will allow them to move forward in a way that makes sense, in a way that they can understand where they're clearly onside and where they're clearly offside.
Good stuff. No, I appreciate it. Transparency, traceability, visibility, and what you're looking for in your vendor, and being able to make sure you can show where things came from. Because the regulators are going to, whether it's them or like you were saying on the panel, bringing in other experts to dive deep into what your model is actually doing and where it's sourced the data from are all things to continue doing if you've been doing it for a long time and your other functions.
And so, Andrew, it has been awesome having you on the program for this micro podcast. Thank you again for joining me. Thank you very much. Appreciate it.
Thank you all for listening today. To learn more about Sellent, visit sellent.com or connect with Andrew on LinkedIn. If you enjoyed this podcast, be sure to check out all of our other episodes and follow us on Apple Podcasts, Spotify, and by visiting duckreek.
com slash podcasts. I'm Rob Stavitsky, and we'll see you in the next episode.
Other episodes covering the same guests and topics, from across The B2B Podcast Index.
Utilizing AI internally to iterate faster and empower smaller teams to upskill w/ Vivek Raghunathan #263The Engineering Leadership Podcast · on generative AI96 / 100
AI Is Ready for Government. Is Government Ready?The So What from BCG · on generative AI84 / 100
Episode 111: Building Your Defences Against AI MisinformationValue Driven Data Science · on AI hallucinations83 / 100
How B2B Marketers Use AI to Personalize at Scale for EnterpriseB2B Marketing with Fexingo · on Human-in-the-loop review82 / 100
How CX Leaders Can Win With AI with Greg MeliaBusiness Transformation Pitch with The CX Goalkeeper · on generative AI77 / 100
RBC's Kelly Bradley on Disclosure, Disruption, and Developing TalentHBS Managing the Future of Work · on generative AI77 / 100Smart Conversations: The Impact of Chatbots in Insurance with Meredith Barnes-Cook of Resource ProThe Importance of Payments for Insurers with Bernhard Schneider of PwCInsurance Accelerators, MGAs, and AI, with David Gritz of Insurtech NYMillennial Perspectives: Forecasting the Insurance Landscape in '24 with Ema RoloffHarnessing Data for Better Insurance Decisions with Sully McConnell of Snowflake