1 00:00:00,080 --> 00:00:02,159 Hello, everyone. This is Jacob Emerson with the 2 00:00:02,159 --> 00:00:03,859 Becker's Payer Issues podcast. 3 00:00:04,240 --> 00:00:06,559 Thrilled today to be joined by doctor Nathan 4 00:00:06,559 --> 00:00:09,139 Kronk, who is the manager of AI engineering 5 00:00:09,199 --> 00:00:09,859 and operations 6 00:00:10,160 --> 00:00:12,719 at Premera Blue Cross. Nathan, thanks so much 7 00:00:12,719 --> 00:00:14,080 for taking the time to be with me 8 00:00:14,080 --> 00:00:16,274 on the podcast today. My pleasure. Thanks for 9 00:00:16,274 --> 00:00:18,315 having me. Yeah. And before we dive into 10 00:00:18,315 --> 00:00:19,835 everything we wanna talk with you about in 11 00:00:19,835 --> 00:00:22,315 a special new project that's been happening at 12 00:00:22,315 --> 00:00:24,394 Premera, can you tell us a little bit 13 00:00:24,394 --> 00:00:26,394 more about yourself and what it is that 14 00:00:26,394 --> 00:00:27,614 you do at the company? 15 00:00:28,719 --> 00:00:29,619 Sure. Yeah. 16 00:00:30,399 --> 00:00:32,000 So I have a my background is in 17 00:00:32,000 --> 00:00:34,179 academia. You know, I I I was researching 18 00:00:34,239 --> 00:00:34,739 AI, 19 00:00:35,200 --> 00:00:37,600 you know, back in 2016 for my PhD, 20 00:00:37,600 --> 00:00:38,100 and, 21 00:00:38,559 --> 00:00:40,239 it kinda sort of taken off, making some 22 00:00:40,239 --> 00:00:41,759 big splash out in the world. And I 23 00:00:41,759 --> 00:00:43,895 thought, you know, maybe I'd like to get 24 00:00:43,895 --> 00:00:45,975 out of the, the the rubber room, the 25 00:00:45,975 --> 00:00:48,215 ivory tower of the the the crazy space 26 00:00:48,215 --> 00:00:50,135 of the academics and go make an impact 27 00:00:50,135 --> 00:00:51,734 with some of this stuff. And health care 28 00:00:51,734 --> 00:00:52,715 seemed like a really, 29 00:00:53,335 --> 00:00:56,075 you know, it transcends race, gender, ethnicity, politics. 30 00:00:56,135 --> 00:00:57,655 Everybody needs health, so it seemed like a 31 00:00:57,655 --> 00:00:59,056 nice place to go try and make a 32 00:00:59,056 --> 00:00:59,609 contribution. So 33 00:01:00,469 --> 00:01:02,710 I joined, Premera, you know, about three years 34 00:01:02,710 --> 00:01:03,210 ago, 35 00:01:03,590 --> 00:01:06,469 and I've had the really enjoying opportunity like, 36 00:01:06,469 --> 00:01:09,269 the enjoyable opportunity to help kind of encourage 37 00:01:09,269 --> 00:01:12,329 and engender this this, adoption of AI technologies 38 00:01:12,469 --> 00:01:14,564 by I go do, like, AI roadshows around 39 00:01:14,564 --> 00:01:15,545 the company, presentations, 40 00:01:15,844 --> 00:01:18,165 you know, on jailbreak attacks, you know, trying 41 00:01:18,165 --> 00:01:21,145 to help people become familiar with, the capabilities 42 00:01:21,204 --> 00:01:23,365 of this technology, building prototypes to show the 43 00:01:23,365 --> 00:01:24,344 art of the possible. 44 00:01:24,805 --> 00:01:26,805 It's been a really fun role that I've 45 00:01:26,805 --> 00:01:27,810 had had here at the company. 46 00:01:28,189 --> 00:01:30,590 But since then, we've actually moved from, you 47 00:01:30,590 --> 00:01:33,070 know, roadshows in education into production. We've got 48 00:01:33,070 --> 00:01:34,210 some pretty impressive, 49 00:01:34,829 --> 00:01:36,990 governance standards in place. We've got, you know, 50 00:01:36,990 --> 00:01:38,750 an AI security gateway that's, 51 00:01:39,125 --> 00:01:41,604 pretty cutting edge and some real Gen AI 52 00:01:41,604 --> 00:01:43,525 products in production making difference in the health 53 00:01:43,525 --> 00:01:44,504 care space. So, 54 00:01:44,805 --> 00:01:45,864 that's a bit about me. 55 00:01:46,405 --> 00:01:48,325 Fantastic. Well, like I said, we really appreciate 56 00:01:48,325 --> 00:01:50,564 you taking the time to be with us 57 00:01:50,564 --> 00:01:51,064 today. 58 00:01:51,520 --> 00:01:53,620 And let's talk about some of those generative 59 00:01:53,760 --> 00:01:55,520 AI products that have been happening at the 60 00:01:55,520 --> 00:01:56,020 company. 61 00:01:56,560 --> 00:01:59,939 There's a new generative AI chatbot called Alice. 62 00:02:00,079 --> 00:02:02,560 So can you walk us through what exactly 63 00:02:02,560 --> 00:02:04,814 that is, how it was developed, and some 64 00:02:04,814 --> 00:02:07,615 of the timeline from concept to the actual 65 00:02:07,615 --> 00:02:09,694 rollouts? Ultimately, can you give us that sense 66 00:02:09,694 --> 00:02:12,094 of of scale in terms of what what's 67 00:02:12,094 --> 00:02:13,794 happening with this at the company? 68 00:02:14,655 --> 00:02:16,754 Yeah. There are some really important details, 69 00:02:17,099 --> 00:02:19,419 to make to make these timelines make sense. 70 00:02:19,419 --> 00:02:20,000 Right? So 71 00:02:20,860 --> 00:02:22,699 about two and a half years ago, we 72 00:02:22,699 --> 00:02:23,099 saw 73 00:02:23,580 --> 00:02:25,819 I mean, you know, my myself and my 74 00:02:25,819 --> 00:02:27,500 team, we came from academia. So we kinda 75 00:02:27,500 --> 00:02:29,020 knew what the cutting edge tech was and 76 00:02:29,020 --> 00:02:30,780 what the art of the possible was. And 77 00:02:30,780 --> 00:02:31,840 we saw this agentic 78 00:02:32,235 --> 00:02:34,555 future coming, or at least we believed that 79 00:02:34,555 --> 00:02:36,014 it was coming, and we started, 80 00:02:36,474 --> 00:02:38,555 building an infrastructure very early on. So some 81 00:02:38,555 --> 00:02:40,235 of the timelines and things I'll share with 82 00:02:40,235 --> 00:02:42,155 you sound pretty quick, but I wanna, you 83 00:02:42,155 --> 00:02:44,634 know, couch them in the the reality, which 84 00:02:44,634 --> 00:02:46,147 is that we started building an AI infrastructure, 85 00:02:46,939 --> 00:02:48,960 for the company that was reusable and reproducible 86 00:02:49,099 --> 00:02:50,159 about two years ago. 87 00:02:50,460 --> 00:02:52,540 So but the the reality is is the 88 00:02:52,540 --> 00:02:55,040 CSR team, you know, they had, the Dynamics 89 00:02:55,580 --> 00:02:56,879 three sixty five tool, 90 00:02:57,419 --> 00:02:59,340 advertised to them as a nice tool to 91 00:02:59,340 --> 00:03:02,159 help support the CSRs in finding information quicker. 92 00:03:02,794 --> 00:03:04,794 And they they went and tried it out, 93 00:03:04,794 --> 00:03:06,175 and it unfortunately, 94 00:03:06,875 --> 00:03:09,754 right, very early on, these these generalist AI 95 00:03:09,754 --> 00:03:11,514 solutions that try and, like, boil the ocean 96 00:03:11,514 --> 00:03:12,175 for everyone, 97 00:03:12,555 --> 00:03:15,275 they they can they can work, but not 98 00:03:15,275 --> 00:03:17,969 fantastically. Right? I mean, when you try and 99 00:03:17,969 --> 00:03:18,949 have a generalized, 100 00:03:19,250 --> 00:03:21,810 you know, the vector database, right, which basically 101 00:03:21,810 --> 00:03:23,969 takes any type of data of any form 102 00:03:23,969 --> 00:03:26,150 and chunks it, partitions it, makes it searchable 103 00:03:26,210 --> 00:03:27,750 for your AI agents, 104 00:03:28,210 --> 00:03:30,689 you're losing a lot of domain expertise that 105 00:03:30,689 --> 00:03:31,510 can be exploited, 106 00:03:32,034 --> 00:03:34,275 by people who really know what the application 107 00:03:34,275 --> 00:03:36,435 is is for. So that was where we 108 00:03:36,435 --> 00:03:38,194 came in. They they came to us and 109 00:03:38,194 --> 00:03:39,794 said, hey. We're trying this Dynamics thing. It 110 00:03:39,794 --> 00:03:42,114 isn't quite working. Why why not? And we 111 00:03:42,114 --> 00:03:44,759 said, well, it's not because there's a lot 112 00:03:44,759 --> 00:03:47,400 of idiosyncratic domain expertise that needs to be 113 00:03:47,400 --> 00:03:50,360 baked into the chunking parse processes and and 114 00:03:50,360 --> 00:03:51,959 how the text is formatted and how the 115 00:03:51,959 --> 00:03:54,520 agent reasons over it. Let's let's show you. 116 00:03:54,520 --> 00:03:56,919 And they said, okay. So, honestly, it took 117 00:03:56,919 --> 00:03:58,675 us about a week to make a prototype 118 00:03:58,675 --> 00:04:00,354 given that we had the infrastructure and everything 119 00:04:00,354 --> 00:04:02,935 in place. And we had a, a reasoning 120 00:04:03,074 --> 00:04:05,395 agent that was built on a technology called 121 00:04:05,395 --> 00:04:06,455 the LLM compiler, 122 00:04:06,834 --> 00:04:09,074 and it searched through thousands of methods and 123 00:04:09,074 --> 00:04:11,074 procedures and would give you answers, would reason 124 00:04:11,074 --> 00:04:12,594 over them and give you answers. And we 125 00:04:12,594 --> 00:04:14,250 said, this is the kind of thing, that 126 00:04:14,250 --> 00:04:16,490 we can do. And they said, great. Let's 127 00:04:16,490 --> 00:04:18,089 let's build this out and go into production 128 00:04:18,089 --> 00:04:18,810 with it. So, 129 00:04:19,290 --> 00:04:21,610 it took us about three months to get 130 00:04:21,610 --> 00:04:22,110 into, 131 00:04:23,129 --> 00:04:25,129 product like, the UAT phase. So that was 132 00:04:25,129 --> 00:04:26,264 when we rolled it out to a couple 133 00:04:26,264 --> 00:04:28,365 of different lines of business to start experimenting 134 00:04:28,425 --> 00:04:30,345 in a small in some small ways. And 135 00:04:30,345 --> 00:04:32,425 then over the past about two to three 136 00:04:32,425 --> 00:04:34,105 months, we've been slowly rolling it out to 137 00:04:34,105 --> 00:04:35,625 more and more, and, you know, we're up 138 00:04:35,625 --> 00:04:37,884 to hundreds of CSRs that are now using 139 00:04:37,944 --> 00:04:39,705 using this tool to help them find methods 140 00:04:39,705 --> 00:04:42,060 and procedures information, more quickly to help members 141 00:04:42,139 --> 00:04:43,519 and and people who call in. 142 00:04:43,980 --> 00:04:45,899 Wow. No. I mean, that's amazing. So a 143 00:04:45,899 --> 00:04:47,500 lot of time went into this, but in 144 00:04:47,500 --> 00:04:49,180 terms of the infrastructure, but then it sounds 145 00:04:49,180 --> 00:04:50,379 like it really a lot of it came 146 00:04:50,379 --> 00:04:53,439 together quickly, once you started developing that prototype. 147 00:04:53,980 --> 00:04:56,139 You mentioned, you know, this is for customer 148 00:04:56,139 --> 00:04:57,040 service operations. 149 00:04:57,985 --> 00:04:59,185 So can you give us a sense of 150 00:04:59,185 --> 00:05:00,785 some of the impact you've seen so far, 151 00:05:00,785 --> 00:05:02,245 what you've heard on the ground, 152 00:05:02,704 --> 00:05:05,044 from from people actually using this tool? 153 00:05:05,584 --> 00:05:07,904 Yeah. That's, that's some of the most rewarding 154 00:05:07,904 --> 00:05:09,264 part of it here. I mean, you know, 155 00:05:09,264 --> 00:05:10,785 we are we love the AI and the 156 00:05:10,785 --> 00:05:12,810 building stuff, but that's you know, hearing the 157 00:05:12,810 --> 00:05:14,589 impact on the front lines has been great. 158 00:05:14,730 --> 00:05:16,410 Yeah. There's there's so many quotes that we've 159 00:05:16,410 --> 00:05:19,470 heard. Firstly, you know, just allowing the CSRs 160 00:05:19,689 --> 00:05:22,569 to hear that the members are happier, like, 161 00:05:22,569 --> 00:05:25,414 on average because they can get information faster 162 00:05:25,414 --> 00:05:26,935 so that they don't sit there and get, 163 00:05:26,935 --> 00:05:28,774 you know, frustrated on the phone. Well, please 164 00:05:28,774 --> 00:05:29,974 give me a minute to go look this 165 00:05:29,974 --> 00:05:32,314 up. And, you know, they're they're calling in 166 00:05:32,615 --> 00:05:34,615 on their lunch breaks or or maybe, you 167 00:05:34,615 --> 00:05:36,534 know, in a break in work. They don't 168 00:05:36,534 --> 00:05:37,814 have a lot of time. That that's a 169 00:05:37,814 --> 00:05:39,360 bit of a luxury. Right? So, 170 00:05:39,919 --> 00:05:41,279 for the CSRs to be able to get 171 00:05:41,279 --> 00:05:43,360 them answers faster has been, 172 00:05:43,839 --> 00:05:45,519 rewarding both for the members and then also 173 00:05:45,519 --> 00:05:47,599 the CSRs to give them more job satisfaction, 174 00:05:47,599 --> 00:05:49,219 which is which has been pretty great. 175 00:05:49,519 --> 00:05:51,759 The the leaders are also quite happy because 176 00:05:51,759 --> 00:05:54,579 we're seeing, trending down in call handle times, 177 00:05:54,774 --> 00:05:55,435 which is, 178 00:05:55,894 --> 00:05:57,654 you know, as as leaders are are well 179 00:05:57,654 --> 00:05:59,675 aware, that's a pretty important metric for, 180 00:06:00,055 --> 00:06:02,454 success in call centers. So average handle time 181 00:06:02,454 --> 00:06:03,035 has been, 182 00:06:03,495 --> 00:06:05,495 trending down as more and more adoption of 183 00:06:05,495 --> 00:06:06,875 the the tool is used. 184 00:06:07,699 --> 00:06:08,600 And in addition, 185 00:06:08,900 --> 00:06:10,819 something else that's pretty neat is we have, 186 00:06:11,060 --> 00:06:13,460 the c the TSSs themselves, which are the 187 00:06:13,460 --> 00:06:16,740 technical support specialists that help CSRs when when 188 00:06:16,740 --> 00:06:18,120 calls get, tricky. 189 00:06:18,580 --> 00:06:20,520 Usually, they're inundated with questions. 190 00:06:20,834 --> 00:06:22,055 But now that the CSRs 191 00:06:22,514 --> 00:06:25,475 are able to go to Alice for some 192 00:06:25,475 --> 00:06:28,274 first level triaging of of challenging questions, the 193 00:06:28,274 --> 00:06:30,194 TSSs now have a bit more time to 194 00:06:30,194 --> 00:06:31,814 go actually improve the documentation 195 00:06:32,355 --> 00:06:34,550 that Alice is using to give the CSRs 196 00:06:34,610 --> 00:06:36,769 the answers, which has been great because, I 197 00:06:36,769 --> 00:06:38,930 mean, there's thousands of documents and they're they've 198 00:06:38,930 --> 00:06:40,930 been maintained by humans. So as you know, 199 00:06:40,930 --> 00:06:42,370 there's there's, of course, gonna be a few 200 00:06:42,370 --> 00:06:44,610 errors and mistakes in them. So Alice has 201 00:06:44,610 --> 00:06:46,769 been surfacing those where maybe Alice gives a 202 00:06:46,769 --> 00:06:49,074 quote, unquote wrong answer. And they go, well, 203 00:06:49,074 --> 00:06:50,754 Alice made a mistake, and they dig in. 204 00:06:50,754 --> 00:06:52,754 And they're like, no. Actually, our our policies 205 00:06:52,754 --> 00:06:54,754 or things have have evolved or changed over 206 00:06:54,754 --> 00:06:56,514 time, so now we can update this. And 207 00:06:56,514 --> 00:06:58,855 it's just making everything better where the documentation 208 00:06:58,995 --> 00:07:01,540 is more aligned more quickly, CSRs get better 209 00:07:01,540 --> 00:07:03,379 answers more quickly, and the leaders have been 210 00:07:03,379 --> 00:07:05,879 pretty pleased with it too. So, it's been 211 00:07:05,939 --> 00:07:08,100 a a pretty rewarding, turnaround to see how 212 00:07:08,100 --> 00:07:09,319 people have been adopting it. 213 00:07:09,860 --> 00:07:11,620 Absolutely. So, really, this sounds like a win 214 00:07:11,620 --> 00:07:13,540 win dynamic both from the member and the 215 00:07:13,540 --> 00:07:14,040 internal 216 00:07:14,785 --> 00:07:17,025 perspective, especially with that average handle time trending 217 00:07:17,025 --> 00:07:19,985 down, the technical support burden, being lifted so 218 00:07:19,985 --> 00:07:22,225 they can focus on other things. For the 219 00:07:22,225 --> 00:07:24,225 other health plan leaders that are listening in, 220 00:07:24,225 --> 00:07:24,725 Nathan, 221 00:07:25,264 --> 00:07:27,345 from that operational and that ROI perspective, I 222 00:07:27,345 --> 00:07:29,125 know you touched on a few metrics here. 223 00:07:29,269 --> 00:07:30,870 Are are there any lessons you would share 224 00:07:30,870 --> 00:07:33,350 in terms of scaling an in house AI 225 00:07:33,350 --> 00:07:36,069 solution, that could that they could apply to, 226 00:07:36,709 --> 00:07:38,170 their companies all over the country? 227 00:07:38,790 --> 00:07:40,790 Yeah. That's a that's a really important question, 228 00:07:40,790 --> 00:07:42,404 and I think we're at a very critical 229 00:07:42,404 --> 00:07:44,084 point where these decisions need to be made 230 00:07:44,084 --> 00:07:45,524 now and they need to meet be made 231 00:07:45,524 --> 00:07:46,504 correctly. Right? 232 00:07:46,884 --> 00:07:47,384 So 233 00:07:47,845 --> 00:07:50,884 everybody's got this build versus buy idea, and 234 00:07:50,884 --> 00:07:53,444 it seems attractive. Right? Well, why would I 235 00:07:53,444 --> 00:07:55,305 try and get an in house AI engineering 236 00:07:55,365 --> 00:07:57,384 team to build these things out when 237 00:07:57,790 --> 00:08:00,189 pretty much every company out there now is, 238 00:08:00,189 --> 00:08:01,330 you know, advertising 239 00:08:01,709 --> 00:08:03,949 AI AI everything. You can buy AI powered 240 00:08:03,949 --> 00:08:06,189 pizza on the way home. Right? Well, of 241 00:08:06,189 --> 00:08:08,990 course, you know, everybody's gonna advertise that because 242 00:08:08,990 --> 00:08:11,230 that's the new attractive hot topic that people, 243 00:08:11,230 --> 00:08:13,125 you know, wanna wanna market. But what I 244 00:08:13,125 --> 00:08:15,625 would encourage leaders to acknowledge is the following. 245 00:08:16,564 --> 00:08:17,865 Most of businesses 246 00:08:19,285 --> 00:08:22,745 is institutional knowledge. Right? Your your real intellectual 247 00:08:22,884 --> 00:08:24,664 property and value is in 248 00:08:25,000 --> 00:08:27,000 oftentimes, your data, of course, but all the 249 00:08:27,000 --> 00:08:28,060 knowledge and expertise 250 00:08:28,519 --> 00:08:30,600 that your staff and your employees bring, all 251 00:08:30,600 --> 00:08:33,980 the processes and and procedures that you've distilled 252 00:08:34,039 --> 00:08:34,860 over decades 253 00:08:35,320 --> 00:08:36,220 of of honing 254 00:08:36,519 --> 00:08:39,100 all this this knowledge and expertise. So 255 00:08:39,615 --> 00:08:40,914 if you can build 256 00:08:41,215 --> 00:08:43,315 AI agents internally that crystallize 257 00:08:44,014 --> 00:08:46,254 all of that expertise that you've owned over 258 00:08:46,254 --> 00:08:46,754 time, 259 00:08:47,134 --> 00:08:47,955 you're basically 260 00:08:48,495 --> 00:08:50,595 locking down that intellectual property. 261 00:08:50,894 --> 00:08:53,534 And that's now an asset for your your 262 00:08:53,534 --> 00:08:54,034 enterprise, 263 00:08:54,559 --> 00:08:56,240 And you can evolve it, grow it over 264 00:08:56,240 --> 00:08:58,720 time, allow those agents to cooperate with other 265 00:08:58,720 --> 00:09:00,799 agents to solve more complex problems as we 266 00:09:00,799 --> 00:09:01,620 start emerging, 267 00:09:02,799 --> 00:09:05,620 evolving towards these emerging, like, multi agent systems. 268 00:09:06,559 --> 00:09:07,059 So 269 00:09:07,684 --> 00:09:10,084 my advice is early on, if possible, if 270 00:09:10,084 --> 00:09:12,164 it makes sense for you, really do try 271 00:09:12,164 --> 00:09:14,105 to to to think through if you can 272 00:09:14,404 --> 00:09:15,784 try to build your own custom, 273 00:09:16,245 --> 00:09:18,485 agents or at a minimum ensure that you 274 00:09:18,485 --> 00:09:20,804 have full IP over the agents that you 275 00:09:20,804 --> 00:09:22,709 are trying to build internally. Yeah. That that's 276 00:09:22,709 --> 00:09:24,789 my advice. Sure. It makes a lot of 277 00:09:24,789 --> 00:09:26,230 sense, and I I think that's great advice 278 00:09:26,230 --> 00:09:28,230 for our audience as well. The the last 279 00:09:28,230 --> 00:09:29,909 thing I wanted to ask you, Nathan, is 280 00:09:29,909 --> 00:09:33,209 how you see AI enabled customer service tools 281 00:09:33,509 --> 00:09:35,110 like the one you've built. How is that 282 00:09:35,110 --> 00:09:37,190 gonna continue to shape Premera if you look 283 00:09:37,190 --> 00:09:39,214 forward the next three to five years? I 284 00:09:39,214 --> 00:09:42,095 know this is very, quickly evolving technology, and 285 00:09:42,095 --> 00:09:43,714 that is a long time frame. But, 286 00:09:44,254 --> 00:09:46,334 what's your predictions here for for both your 287 00:09:46,334 --> 00:09:48,674 company but also for the wider industry? 288 00:09:49,534 --> 00:09:51,214 Yeah. I mean, I can I can tell 289 00:09:51,214 --> 00:09:52,355 you? Right? So, I mean, we 290 00:09:52,789 --> 00:09:55,850 have across all of our organizations, right, cybersecurity, 291 00:09:56,230 --> 00:09:58,009 IT, corporate data and analytics, 292 00:09:58,389 --> 00:09:59,669 you know, we've been coming together, and we've 293 00:09:59,669 --> 00:10:01,209 got a core group of people 294 00:10:01,669 --> 00:10:03,769 that have been making, you know, our enterprise 295 00:10:03,829 --> 00:10:06,230 AI architecture and trying to ensure that our 296 00:10:06,389 --> 00:10:07,875 we're future proofing our adoption 297 00:10:09,075 --> 00:10:10,675 of AI and where we're scaling towards. So 298 00:10:10,675 --> 00:10:12,615 the main thing to acknowledge is that, you 299 00:10:13,235 --> 00:10:14,534 know, health health insurance 300 00:10:15,154 --> 00:10:17,595 I mean, it's it's a math problem at 301 00:10:17,595 --> 00:10:19,075 at the end of the day. Right? So 302 00:10:19,075 --> 00:10:21,394 first and foremost is you gotta solve the 303 00:10:21,394 --> 00:10:23,580 math problem. Computers are better at solving that 304 00:10:23,580 --> 00:10:25,820 math problem, and we wanna acknowledge that that 305 00:10:25,820 --> 00:10:27,420 is the core capability that we want, you 306 00:10:27,420 --> 00:10:29,820 know, technology to support in. But then also 307 00:10:29,820 --> 00:10:31,500 there's the human element. I mean, health is 308 00:10:31,500 --> 00:10:33,340 really important, and we need to be able 309 00:10:33,340 --> 00:10:34,960 to reach out to people, be sensitive, 310 00:10:35,500 --> 00:10:37,040 and be able to connect with people 311 00:10:37,495 --> 00:10:40,154 as well. So what we're thinking is that 312 00:10:40,694 --> 00:10:43,595 this insurance is basically a process driven industry. 313 00:10:43,814 --> 00:10:45,194 So the vast majority 314 00:10:45,495 --> 00:10:47,254 of the the math problems that are being 315 00:10:47,254 --> 00:10:48,414 solved and the input and processing of documents 316 00:10:48,414 --> 00:10:49,115 and output of 317 00:10:51,059 --> 00:10:53,540 documents. This these are things that AI agents 318 00:10:53,540 --> 00:10:55,860 can help automate. And what we would then 319 00:10:55,860 --> 00:10:57,139 be able to do is to leverage all 320 00:10:57,139 --> 00:10:59,059 of the human expertise to to really touch 321 00:10:59,059 --> 00:11:00,500 on the human side of things and make 322 00:11:00,500 --> 00:11:01,720 sure that all the members, 323 00:11:02,144 --> 00:11:04,225 are really feeling cared for or that they 324 00:11:04,225 --> 00:11:06,545 have the support that they really need, so 325 00:11:06,545 --> 00:11:09,524 we can automate the laborious slow sort of 326 00:11:09,825 --> 00:11:13,105 procedural factory line type work, and allow humans 327 00:11:13,105 --> 00:11:15,024 to connect more on the health element, which 328 00:11:15,024 --> 00:11:16,225 is, you know, I think at the end 329 00:11:16,225 --> 00:11:17,710 of the day what what we really want. 330 00:11:17,710 --> 00:11:19,309 That that that's one of the things that 331 00:11:19,309 --> 00:11:21,470 is gonna separate the health industry from others 332 00:11:21,470 --> 00:11:23,870 in AI adoption is that it's very personal 333 00:11:23,870 --> 00:11:26,429 and it's very, very important. And so how 334 00:11:26,429 --> 00:11:28,830 AI is integrated there is gonna be, a 335 00:11:28,830 --> 00:11:29,570 pretty important 336 00:11:29,950 --> 00:11:31,629 thing that these companies need to think through 337 00:11:31,629 --> 00:11:32,129 carefully. 338 00:11:32,764 --> 00:11:34,845 Wonderful. Well, Nathan, I wanna thank you so 339 00:11:34,845 --> 00:11:36,845 much for taking the time to chat with 340 00:11:36,845 --> 00:11:38,524 us and with our audience and for sharing 341 00:11:38,524 --> 00:11:40,845 our insights with all the leaders listening in. 342 00:11:40,845 --> 00:11:43,136 We we really appreciate it. It's my pleasure. 343 00:11:43,136 --> 00:11:44,816 Thank you for having me. Yeah. And and 344 00:11:44,816 --> 00:11:46,496 to our listeners, if you'd like to listen 345 00:11:46,496 --> 00:11:48,656 to more podcasts from Becker's Health Care, you 346 00:11:48,656 --> 00:11:51,316 can visit beckershospitalreview.com.