1 00:00:00,080 --> 00:00:02,560 Hello, everyone. This is Erica Spicer Mason with 2 00:00:02,560 --> 00:00:04,879 Becker's Healthcare. Thank you so much for tuning 3 00:00:04,879 --> 00:00:07,219 in to the Becker's Healthcare podcast series. 4 00:00:07,679 --> 00:00:09,839 So today, we're going to discuss AI and 5 00:00:09,839 --> 00:00:12,500 the payer landscape and how to balance innovation, 6 00:00:12,639 --> 00:00:13,859 ethics, and regulatory 7 00:00:14,160 --> 00:00:14,660 change. 8 00:00:14,974 --> 00:00:17,315 And joining me for this discussion is Nitu 9 00:00:17,375 --> 00:00:17,875 Rajpal, 10 00:00:18,175 --> 00:00:20,275 the CEO of Lilac Software. 11 00:00:20,975 --> 00:00:23,054 Nitu, welcome to the podcast. Thank you so 12 00:00:23,054 --> 00:00:25,214 much for making time for us today. Thank 13 00:00:25,214 --> 00:00:27,570 you, Erica. It's really great to be here. 14 00:00:27,969 --> 00:00:29,890 Oh, we're so thrilled to have you. And 15 00:00:29,890 --> 00:00:32,130 before we get into our discussion, wanted to 16 00:00:32,130 --> 00:00:33,489 see if you'd like to share a little 17 00:00:33,489 --> 00:00:35,329 bit more about yourself, your work in health 18 00:00:35,329 --> 00:00:36,629 care, your organization, 19 00:00:37,009 --> 00:00:38,950 whatever feels pertinent to listeners. 20 00:00:39,649 --> 00:00:43,045 Oh, thank you. So I'm Neetu Rajpal, and 21 00:00:43,045 --> 00:00:43,864 I am the 22 00:00:44,324 --> 00:00:46,905 CEO and cofounder of Lilac Software. 23 00:00:47,604 --> 00:00:51,204 With Lilac Software, we're very focused on making 24 00:00:51,204 --> 00:00:53,880 sure good data gets to all the practitioners 25 00:00:54,260 --> 00:00:57,079 that are currently in health care. We're building 26 00:00:57,140 --> 00:00:58,600 a fully managed 27 00:00:59,140 --> 00:01:01,399 solution for data analytics, 28 00:01:01,780 --> 00:01:03,079 insights, AI 29 00:01:03,620 --> 00:01:05,159 for payers and providers. 30 00:01:05,805 --> 00:01:07,664 Before my work at Lilac, 31 00:01:08,524 --> 00:01:11,664 I was the CTO of Oscar Health. I 32 00:01:11,724 --> 00:01:14,784 appreciated and loved being at Oscar Health. 33 00:01:15,244 --> 00:01:17,760 I oversaw work for all of the tech 34 00:01:17,840 --> 00:01:20,099 stack that's used to run the insurance company, 35 00:01:20,159 --> 00:01:21,380 everything from 36 00:01:21,760 --> 00:01:25,840 from claims processing to enrollment, billing, brokers, member 37 00:01:25,840 --> 00:01:29,359 engagement. And going through that journey really, like, 38 00:01:29,359 --> 00:01:32,020 inspired what I am doing at EyeLac now. 39 00:01:32,275 --> 00:01:34,194 And and overall, my career has all been 40 00:01:34,194 --> 00:01:36,354 building software. So I've been building software for 41 00:01:36,354 --> 00:01:37,334 multiple decades, 42 00:01:37,875 --> 00:01:41,314 and enterprise data analytics is is my passion 43 00:01:41,314 --> 00:01:42,994 to get make sure the right data gets 44 00:01:42,994 --> 00:01:45,349 to the right people to make good decisions 45 00:01:45,409 --> 00:01:48,390 and to help them with, with agents. 46 00:01:49,649 --> 00:01:52,149 Yeah. Me too. I really appreciate the background 47 00:01:52,289 --> 00:01:54,369 information. And it it sounds like you have 48 00:01:54,369 --> 00:01:56,209 the perfect blend of experience, you know, coming 49 00:01:56,209 --> 00:01:58,289 from a a major payer organization and all 50 00:01:58,289 --> 00:02:00,394 of your decades of experience in software development 51 00:02:00,394 --> 00:02:01,134 and building. 52 00:02:01,594 --> 00:02:03,274 Sounds like you are the perfect person to 53 00:02:03,274 --> 00:02:04,334 talk to us about 54 00:02:04,715 --> 00:02:06,174 AI and the payer landscape 55 00:02:06,474 --> 00:02:07,454 today. So 56 00:02:08,074 --> 00:02:10,014 I wanted to start us off by 57 00:02:10,314 --> 00:02:12,689 just kind of acknowledging a trend that I 58 00:02:12,689 --> 00:02:14,530 know I've seen and our audience has seen 59 00:02:14,530 --> 00:02:17,330 in health care coverage lately, and that's the 60 00:02:17,330 --> 00:02:19,909 fact that Agentic AI is really gaining momentum. 61 00:02:20,530 --> 00:02:23,009 So can you provide a brief overview of 62 00:02:23,009 --> 00:02:25,330 what this technology is and maybe even share 63 00:02:25,330 --> 00:02:27,330 a practical example of how it's being used 64 00:02:27,330 --> 00:02:27,989 right now? 65 00:02:28,644 --> 00:02:31,604 Yeah. I think Adjentic AI is one of 66 00:02:31,604 --> 00:02:35,044 the most exciting advancements in in in AI. 67 00:02:35,044 --> 00:02:35,544 I 68 00:02:35,844 --> 00:02:38,084 I know I say that in addition to 69 00:02:38,084 --> 00:02:40,324 the fact that there was already chat TPT 70 00:02:40,324 --> 00:02:41,705 and there's machine learning. 71 00:02:42,080 --> 00:02:43,360 But just to give a bit of a 72 00:02:43,360 --> 00:02:45,460 primer, machine learning in is 73 00:02:45,840 --> 00:02:46,500 is colloquially 74 00:02:46,800 --> 00:02:48,960 more about, like, finding the right set of 75 00:02:48,960 --> 00:02:51,219 patterns and data and understanding data. 76 00:02:51,840 --> 00:02:54,180 Large language models use that information 77 00:02:54,479 --> 00:02:57,280 to then understand and get context and and 78 00:02:57,280 --> 00:02:57,914 to forecast 79 00:02:59,275 --> 00:03:01,914 build new documents or build new conversations or 80 00:03:01,914 --> 00:03:02,894 build new things. 81 00:03:03,995 --> 00:03:04,495 Agentic 82 00:03:05,114 --> 00:03:06,734 AI is this exciting 83 00:03:07,275 --> 00:03:07,775 advancement 84 00:03:08,074 --> 00:03:08,574 where 85 00:03:08,954 --> 00:03:12,250 Agentic AI systems can, like, really be in 86 00:03:12,250 --> 00:03:14,810 the world and understand it and observe it 87 00:03:14,810 --> 00:03:16,030 and learn from it, 88 00:03:16,330 --> 00:03:18,490 and then reason about it and build a 89 00:03:18,490 --> 00:03:20,730 plan to take action, and then go ahead 90 00:03:20,730 --> 00:03:23,230 and take an action. So that all sounds 91 00:03:23,290 --> 00:03:26,685 very much like like what a human AI 92 00:03:26,685 --> 00:03:29,085 assistant would do, and and I think that's 93 00:03:29,085 --> 00:03:31,564 a pretty good way of thinking about agentic 94 00:03:31,564 --> 00:03:33,105 systems and agentic AI, 95 00:03:34,205 --> 00:03:34,705 assistants 96 00:03:35,085 --> 00:03:36,305 or independently 97 00:03:36,764 --> 00:03:38,844 operating folks who can start doing more and 98 00:03:38,844 --> 00:03:39,504 more things. 99 00:03:39,805 --> 00:03:43,280 And as the assistants get more familiar with 100 00:03:43,280 --> 00:03:45,120 the work that they're doing, they get more 101 00:03:45,120 --> 00:03:46,879 and more independent, and then I think that's 102 00:03:46,879 --> 00:03:48,099 also a good corollary. 103 00:03:48,639 --> 00:03:51,459 In in terms of, like, health care world, 104 00:03:52,239 --> 00:03:54,479 a a simple simple example that I can 105 00:03:54,479 --> 00:03:56,564 come up with is something like, think about, 106 00:03:56,564 --> 00:04:00,104 like, appeals and grievances or grievances against claims. 107 00:04:00,564 --> 00:04:03,525 And it's somebody files a grievance about a 108 00:04:03,525 --> 00:04:04,504 denied claim. 109 00:04:05,044 --> 00:04:06,965 AI, in this case, can step in and 110 00:04:06,965 --> 00:04:10,085 streamline the process of, like, working through that 111 00:04:10,085 --> 00:04:12,969 grievance as in they can cross reference all 112 00:04:12,969 --> 00:04:15,449 of the information that's available about the member, 113 00:04:15,449 --> 00:04:17,949 about the policy and the plan, about identifying 114 00:04:18,009 --> 00:04:20,329 what the discrepancy is, if there's any missing 115 00:04:20,329 --> 00:04:20,829 information. 116 00:04:21,290 --> 00:04:23,610 AI can gather all of this information behind 117 00:04:23,610 --> 00:04:26,004 the scenes without a person being involved, And 118 00:04:26,004 --> 00:04:28,324 then they can, like, understand the context and 119 00:04:28,324 --> 00:04:28,824 then 120 00:04:29,204 --> 00:04:29,704 generate 121 00:04:30,084 --> 00:04:33,064 a personalized response for this particular 122 00:04:33,444 --> 00:04:34,504 claim grievance. 123 00:04:34,884 --> 00:04:36,725 And that, I think, is the most exciting 124 00:04:36,725 --> 00:04:38,884 part. Then they bring a plan and an 125 00:04:38,884 --> 00:04:40,379 execution plan perhaps 126 00:04:40,860 --> 00:04:42,879 to to a human who can approve it, 127 00:04:43,420 --> 00:04:45,920 but with all of the supporting information included, 128 00:04:46,300 --> 00:04:49,020 and then go execute on that plan. And 129 00:04:49,020 --> 00:04:52,000 this this, like, super dedicated problem solver 130 00:04:52,379 --> 00:04:54,560 doesn't actually have a scale requirement. 131 00:04:54,904 --> 00:04:57,384 A person can totally do it, and a 132 00:04:57,384 --> 00:04:59,485 person will likely do it in most complicated 133 00:04:59,625 --> 00:05:02,925 scenarios for a long time. But the simpler 134 00:05:03,225 --> 00:05:06,345 scenarios can actually be handled by an agentic 135 00:05:06,345 --> 00:05:08,425 system in this case. So that I think 136 00:05:08,425 --> 00:05:10,740 is a pretty practical example, and I know 137 00:05:10,740 --> 00:05:12,899 of places that are already starting to implement 138 00:05:12,899 --> 00:05:13,399 this. 139 00:05:14,339 --> 00:05:16,419 Fascinating. And I think you hit on an 140 00:05:16,419 --> 00:05:18,980 area that certainly is kind of the center 141 00:05:18,980 --> 00:05:20,660 of a lot of conversations when it comes 142 00:05:20,660 --> 00:05:22,519 to efficiency in health care organizations, 143 00:05:22,819 --> 00:05:25,000 whether it's on the provider or payer side. 144 00:05:25,194 --> 00:05:27,115 You know, claims is is such an area 145 00:05:27,115 --> 00:05:29,274 where folks are really trying to identify those 146 00:05:29,274 --> 00:05:31,455 opportunities to streamline things. So 147 00:05:31,834 --> 00:05:34,475 really appreciate that example. And I was actually 148 00:05:34,475 --> 00:05:36,955 planning to ask you also how how payers 149 00:05:36,955 --> 00:05:39,514 can leverage Genentech AI maybe to even help 150 00:05:39,514 --> 00:05:42,350 address quality or care delivery challenges. Could you 151 00:05:42,350 --> 00:05:44,350 shed some light on that and maybe elaborate 152 00:05:44,350 --> 00:05:46,509 on any improvements you've seen in those use 153 00:05:46,509 --> 00:05:47,009 cases? 154 00:05:47,790 --> 00:05:48,290 Absolutely. 155 00:05:49,310 --> 00:05:51,870 As of right now, like, the the STARS 156 00:05:51,870 --> 00:05:54,735 program and the and the HEDIS gap closure 157 00:05:54,794 --> 00:05:57,754 and just generally performing well on STARS is 158 00:05:57,754 --> 00:05:59,774 very near and dear to my heart. 159 00:06:00,314 --> 00:06:02,314 So I'll speak a pick an example from 160 00:06:02,314 --> 00:06:03,214 there. So 161 00:06:03,595 --> 00:06:04,095 most 162 00:06:04,474 --> 00:06:06,875 folks in in pair organizations that I've talked 163 00:06:06,875 --> 00:06:10,310 to try very, very hard to, quote, unquote, 164 00:06:10,370 --> 00:06:12,930 close every gap as in identifying all of 165 00:06:12,930 --> 00:06:15,350 the members that need a particular service 166 00:06:16,610 --> 00:06:19,089 and help them get that service. It could 167 00:06:19,089 --> 00:06:21,335 be it could be as simple as ensuring 168 00:06:21,395 --> 00:06:23,634 that all the right members who should be 169 00:06:23,634 --> 00:06:25,555 getting a breast cancer screening are, in fact, 170 00:06:25,555 --> 00:06:27,555 getting a breast cancer screening. The way this 171 00:06:27,555 --> 00:06:28,455 happens today 172 00:06:29,154 --> 00:06:30,055 is is 173 00:06:30,754 --> 00:06:33,779 some person generates some list, and then some 174 00:06:33,779 --> 00:06:36,420 other person takes that list, and then some 175 00:06:36,420 --> 00:06:38,420 other person or the same person makes a 176 00:06:38,420 --> 00:06:40,740 phone call and and and then tries to 177 00:06:40,740 --> 00:06:43,220 make another phone call to encourage the member 178 00:06:43,220 --> 00:06:45,699 to to go schedule an appointment with a 179 00:06:45,699 --> 00:06:46,199 provider 180 00:06:46,564 --> 00:06:49,204 and then go to the appointment and and 181 00:06:49,204 --> 00:06:50,904 and, like, have have a mammogram. 182 00:06:52,245 --> 00:06:55,204 All of this is ripe for an agentic 183 00:06:55,204 --> 00:06:57,604 system to do everything behind the scenes as 184 00:06:57,604 --> 00:06:59,764 in identify the members that need a gap 185 00:06:59,764 --> 00:07:00,264 closed, 186 00:07:01,240 --> 00:07:03,399 figure out the right time and way that 187 00:07:03,399 --> 00:07:04,939 that gap needs to be closed, 188 00:07:05,639 --> 00:07:07,740 making a phone call to the member, 189 00:07:08,120 --> 00:07:10,839 identifying that the barrier is actually lack of 190 00:07:10,839 --> 00:07:13,159 scheduling, then maybe just, like, going ahead and 191 00:07:13,159 --> 00:07:14,379 scheduling an appointment. 192 00:07:14,685 --> 00:07:17,084 And then also maybe recognizing there are some, 193 00:07:17,084 --> 00:07:18,144 like, nonmedical 194 00:07:18,524 --> 00:07:19,404 determinants of, 195 00:07:20,285 --> 00:07:22,685 detractors of health here. So maybe the member 196 00:07:22,685 --> 00:07:25,084 can't get to the to their appointment on 197 00:07:25,084 --> 00:07:27,404 time. So, like, booking an Uber as well. 198 00:07:27,404 --> 00:07:29,105 So everything to do from 199 00:07:29,540 --> 00:07:31,399 knowing who needs help, 200 00:07:32,180 --> 00:07:32,680 identifying 201 00:07:33,220 --> 00:07:36,519 how they need help, identifying the the blockers 202 00:07:36,579 --> 00:07:39,160 for them to close that gap, and then 203 00:07:39,300 --> 00:07:41,399 executing on all of those pieces 204 00:07:41,714 --> 00:07:44,035 behind the scenes to just make it easy 205 00:07:44,035 --> 00:07:45,955 for the member to actually go get their 206 00:07:45,955 --> 00:07:46,455 mammogram. 207 00:07:47,154 --> 00:07:49,235 What I think is, like, super exciting and 208 00:07:49,235 --> 00:07:51,654 where it, like, really transforms things 209 00:07:52,115 --> 00:07:54,675 is it is not necessarily that you can 210 00:07:54,675 --> 00:07:56,539 do this for a group of members or 211 00:07:56,539 --> 00:07:58,139 a subset of members. You can, in fact, 212 00:07:58,139 --> 00:08:00,860 do this for every member because you're no 213 00:08:00,860 --> 00:08:03,599 longer limited. Your scale is no longer limited 214 00:08:03,659 --> 00:08:05,740 by the number of people who can make 215 00:08:05,740 --> 00:08:07,419 phone calls or the amount of time a 216 00:08:07,419 --> 00:08:07,740 person 217 00:08:08,379 --> 00:08:09,680 you can scale horizontally 218 00:08:10,139 --> 00:08:10,959 to compute. 219 00:08:11,475 --> 00:08:13,794 You can have lots of agents doing lots 220 00:08:13,794 --> 00:08:16,115 of things all at the same time, and 221 00:08:16,115 --> 00:08:19,314 you can personalize care. You can personalize outreach. 222 00:08:19,314 --> 00:08:20,214 You can personalize 223 00:08:20,995 --> 00:08:23,634 the the the plan the member needs to 224 00:08:23,634 --> 00:08:25,710 go close the care gap. So I'm, like, 225 00:08:25,710 --> 00:08:29,089 super excited about super excited about the possibilities 226 00:08:29,389 --> 00:08:30,210 here where 227 00:08:30,910 --> 00:08:31,410 personalization 228 00:08:31,790 --> 00:08:34,429 of health care is is that is is 229 00:08:34,429 --> 00:08:35,410 available now. 230 00:08:37,070 --> 00:08:38,850 Yeah. Me too. That that example 231 00:08:39,375 --> 00:08:41,615 of the care gap list really hits home 232 00:08:41,615 --> 00:08:43,695 for me personally. I had a past life 233 00:08:43,695 --> 00:08:45,294 working as a health educator at a health 234 00:08:45,294 --> 00:08:47,774 plan, and I remember manually working off of 235 00:08:47,774 --> 00:08:50,095 those lists and doing the phone call outreach 236 00:08:50,095 --> 00:08:50,595 and 237 00:08:51,134 --> 00:08:52,980 talk about time consuming. So to your point 238 00:08:52,980 --> 00:08:54,820 about being able to scale that, not just 239 00:08:54,820 --> 00:08:57,379 to a small subset of members, but potentially 240 00:08:57,379 --> 00:08:59,220 all members at a health plan, it's it's 241 00:08:59,220 --> 00:09:00,440 a really exciting idea. 242 00:09:01,379 --> 00:09:03,879 Yeah. I'm very excited for our future here. 243 00:09:04,340 --> 00:09:04,840 Yeah. 244 00:09:05,144 --> 00:09:05,884 Yeah. And 245 00:09:06,345 --> 00:09:08,764 so a follow-up question for you too. So 246 00:09:09,065 --> 00:09:11,144 it sounds like how health plans are are 247 00:09:11,144 --> 00:09:11,644 certainly 248 00:09:12,105 --> 00:09:14,425 experimenting with AI, piloting it, and even using 249 00:09:14,425 --> 00:09:15,884 it in these ways as you've mentioned. 250 00:09:16,504 --> 00:09:19,379 And as they're as they're innovating here, how 251 00:09:19,379 --> 00:09:21,639 can they also ensure that these AI tools 252 00:09:21,940 --> 00:09:25,220 are ethical and unbiased and also patient focused, 253 00:09:25,220 --> 00:09:27,480 you know, especially when we're talking about Medicare 254 00:09:27,620 --> 00:09:29,160 or Medicare Advantage population? 255 00:09:30,415 --> 00:09:33,855 Yeah. Really, really excellent question. We we all 256 00:09:33,855 --> 00:09:35,934 know that your AI models are only going 257 00:09:35,934 --> 00:09:37,615 to be as good as the data that 258 00:09:37,615 --> 00:09:40,035 they've been used to train get trained on, 259 00:09:40,254 --> 00:09:42,975 and any bias in that data will come 260 00:09:42,975 --> 00:09:46,019 through in the models themselves. So I think 261 00:09:46,019 --> 00:09:48,740 the very least, it it needs to it 262 00:09:48,740 --> 00:09:49,959 needs to be monitored 263 00:09:50,500 --> 00:09:53,139 as to which model you're using to implement 264 00:09:53,139 --> 00:09:55,620 your solution and maybe trying a few different 265 00:09:55,620 --> 00:09:58,579 ones and and having lots of spot checks 266 00:09:58,579 --> 00:09:59,399 in the middle. 267 00:09:59,875 --> 00:10:02,535 At least for a while until these systems, 268 00:10:02,675 --> 00:10:04,855 like, you we can we can eliminate, 269 00:10:06,035 --> 00:10:08,615 known biases from these models. Or 270 00:10:09,154 --> 00:10:11,575 so I think the the mitigating bias 271 00:10:12,329 --> 00:10:14,009 based on the data that the models have 272 00:10:14,009 --> 00:10:15,949 been trained on and mitigating 273 00:10:16,730 --> 00:10:20,269 that risk by using multiple models is is 274 00:10:20,490 --> 00:10:21,549 is the first one. 275 00:10:22,009 --> 00:10:24,644 Secondly, I I would encourage that, like, when 276 00:10:24,644 --> 00:10:26,985 we try things, we actually keep the member 277 00:10:27,125 --> 00:10:29,304 at the center of all of it and, 278 00:10:29,764 --> 00:10:32,644 ensure that, like, the real world needs are 279 00:10:32,644 --> 00:10:33,144 met. 280 00:10:33,605 --> 00:10:36,404 So we're not we're not building tech ahead 281 00:10:36,404 --> 00:10:38,745 of its ability to be used in productive 282 00:10:38,884 --> 00:10:39,384 ways. 283 00:10:39,839 --> 00:10:43,120 And that may not always mean just everything 284 00:10:43,120 --> 00:10:45,279 all tech. It may mean a blended version 285 00:10:45,279 --> 00:10:45,779 of 286 00:10:46,159 --> 00:10:49,279 tech and process. And and one of the 287 00:10:49,279 --> 00:10:51,220 last things I'd say here is 288 00:10:52,524 --> 00:10:53,024 picking 289 00:10:53,804 --> 00:10:56,445 the amount of automation you want here. 290 00:10:56,764 --> 00:10:57,664 I think there 291 00:10:58,044 --> 00:10:59,345 tends to be a lot 292 00:11:00,284 --> 00:11:01,584 of intention to 293 00:11:02,044 --> 00:11:03,264 automate everything. 294 00:11:04,044 --> 00:11:06,125 I have been doing tech for a very, 295 00:11:06,125 --> 00:11:07,985 very long time, and I think 296 00:11:08,579 --> 00:11:09,240 the first 297 00:11:10,179 --> 00:11:12,980 seventy percent is usually the most exciting and 298 00:11:12,980 --> 00:11:14,819 the fastest piece to get to. The next 299 00:11:14,819 --> 00:11:16,419 10 is a little bit harder, and then 300 00:11:16,419 --> 00:11:18,839 the last 20 is really, really hard. 301 00:11:19,220 --> 00:11:20,120 And preciseness 302 00:11:20,579 --> 00:11:21,720 and and accuracy 303 00:11:22,675 --> 00:11:25,735 in health care specifically matters a lot. So 304 00:11:25,955 --> 00:11:28,915 I actually think it's perfectly fine to go 305 00:11:28,915 --> 00:11:32,115 with, like, aim for 80% as long as 306 00:11:32,115 --> 00:11:35,475 the accuracy and and completeness are really high 307 00:11:35,475 --> 00:11:37,500 on the required list of things. 308 00:11:38,700 --> 00:11:42,860 And given how, like, stretched thin almost everybody 309 00:11:42,860 --> 00:11:45,899 in a health plan is, I think 80% 310 00:11:45,899 --> 00:11:46,879 is pretty awesome. 311 00:11:47,500 --> 00:11:50,539 So mitigating bias in the data that's used 312 00:11:50,539 --> 00:11:52,240 by using multiple models, 313 00:11:52,884 --> 00:11:53,384 embedding 314 00:11:53,924 --> 00:11:54,424 patient 315 00:11:55,044 --> 00:11:57,764 or member at the center of whatever value 316 00:11:57,764 --> 00:11:59,865 that we're bringing. And finally, like, 317 00:12:00,485 --> 00:12:02,884 I think it's it's worth worth the question 318 00:12:02,884 --> 00:12:04,565 to ask if % 319 00:12:04,565 --> 00:12:06,084 is really should be the goal. Maybe it 320 00:12:06,084 --> 00:12:08,309 should just be 80% for for a while 321 00:12:08,309 --> 00:12:09,370 until we're sure. 322 00:12:10,549 --> 00:12:13,269 I really appreciate how practical those tips are. 323 00:12:13,269 --> 00:12:15,110 Me too. So thank you for that. And 324 00:12:15,110 --> 00:12:16,329 I think that 80% 325 00:12:16,470 --> 00:12:19,110 metric that you mentioned, it's definitely something interesting 326 00:12:19,110 --> 00:12:21,269 for leaders to keep in mind as they 327 00:12:21,269 --> 00:12:22,490 move forward with innovation. 328 00:12:23,834 --> 00:12:24,575 And similarly, 329 00:12:24,955 --> 00:12:27,034 you know, we're talking about ethics here with 330 00:12:27,034 --> 00:12:28,495 AI and technology implementation. 331 00:12:29,434 --> 00:12:31,355 I I know that we're also seeing in 332 00:12:31,355 --> 00:12:33,214 the news some headlines about 333 00:12:33,595 --> 00:12:34,495 AI regulations 334 00:12:35,115 --> 00:12:37,115 kind of going away or at least being 335 00:12:37,115 --> 00:12:38,174 much less restrictive, 336 00:12:38,679 --> 00:12:41,320 and that's, you know, rapidly changing and evolving 337 00:12:41,320 --> 00:12:42,360 in 2025. 338 00:12:42,360 --> 00:12:45,000 So how do you foresee the role of 339 00:12:45,000 --> 00:12:47,480 AI in health care changing, especially shaping the 340 00:12:47,480 --> 00:12:50,059 future of value based care and Medicare programs? 341 00:12:51,735 --> 00:12:53,194 Yeah. I hope 342 00:12:53,495 --> 00:12:56,454 that we will build systems that are built 343 00:12:56,454 --> 00:12:59,355 on top of systems that are currently subject 344 00:12:59,495 --> 00:13:00,794 to HIPAA and 345 00:13:01,174 --> 00:13:03,595 security and compliance and privacy 346 00:13:03,975 --> 00:13:04,475 and 347 00:13:04,980 --> 00:13:05,480 security 348 00:13:05,940 --> 00:13:08,740 rules around there. And I think a really 349 00:13:08,740 --> 00:13:11,000 good way would be to assume your own 350 00:13:11,139 --> 00:13:13,700 AI system is also subject to the same 351 00:13:13,700 --> 00:13:17,139 rules. So data privacy, especially when it comes 352 00:13:17,139 --> 00:13:20,535 to health care, seems really, really valuable data 353 00:13:20,535 --> 00:13:21,195 to protect 354 00:13:21,735 --> 00:13:22,875 even if the regulations 355 00:13:23,415 --> 00:13:25,654 lag in time. Because I think where there 356 00:13:25,654 --> 00:13:27,274 is a there is a 357 00:13:27,654 --> 00:13:28,154 really 358 00:13:28,934 --> 00:13:29,434 different 359 00:13:31,410 --> 00:13:34,210 regulations and the advancements in AI are moving 360 00:13:34,210 --> 00:13:35,750 at. And and I think 361 00:13:36,529 --> 00:13:39,090 the people's privacy and people's privacy on their 362 00:13:39,090 --> 00:13:41,809 health care data is is paramount, and I 363 00:13:41,809 --> 00:13:43,590 and I really encourage the 364 00:13:44,465 --> 00:13:47,605 plans using AI to impose the same requirement 365 00:13:47,665 --> 00:13:50,545 on themselves and while the regulation is still 366 00:13:50,545 --> 00:13:51,605 being worked out. 367 00:13:52,785 --> 00:13:53,285 Yeah. 368 00:13:53,745 --> 00:13:55,184 I I really appreciate how you put that 369 00:13:55,184 --> 00:13:56,865 need to, you know, how there's kind of 370 00:13:56,865 --> 00:13:59,070 this discrepancy in the velocity of the speed 371 00:13:59,070 --> 00:14:01,050 of innovation and and the regulations 372 00:14:01,509 --> 00:14:02,649 behind it. So 373 00:14:03,269 --> 00:14:05,509 really appreciate all of these insights that you've 374 00:14:05,509 --> 00:14:08,230 shared with me and our listeners today. Is 375 00:14:08,230 --> 00:14:10,070 there anything else that you wanted to highlight 376 00:14:10,070 --> 00:14:11,990 or maybe anything we didn't touch on but 377 00:14:11,990 --> 00:14:12,970 you wanna mention? 378 00:14:14,054 --> 00:14:16,394 Yeah. Thank you for the opportunity. I, 379 00:14:17,175 --> 00:14:19,335 actually wanna just talk. Like, I've I am 380 00:14:19,335 --> 00:14:21,735 a technologist, and I've been a technologist pretty 381 00:14:21,735 --> 00:14:23,115 much my whole career. 382 00:14:23,894 --> 00:14:24,955 One of the 383 00:14:25,509 --> 00:14:28,309 things that I find the most fascinating about 384 00:14:28,309 --> 00:14:30,789 doing things in health care is it is 385 00:14:30,789 --> 00:14:32,490 a very integrated environment, 386 00:14:32,870 --> 00:14:33,449 I think. 387 00:14:33,829 --> 00:14:34,329 So 388 00:14:35,110 --> 00:14:35,929 being able 389 00:14:36,470 --> 00:14:37,610 to bring technology 390 00:14:38,389 --> 00:14:41,154 to a plan, but fully recognizing 391 00:14:41,534 --> 00:14:43,934 the current state of the world is going 392 00:14:43,934 --> 00:14:45,855 to be really helpful. It's a it's a 393 00:14:45,855 --> 00:14:47,694 bit of a a weird way to say 394 00:14:47,694 --> 00:14:49,934 it, but I think bringing everybody along is 395 00:14:49,934 --> 00:14:52,014 actually going to be the fastest way to 396 00:14:52,014 --> 00:14:54,095 get this technology adopted. It's like you how 397 00:14:54,095 --> 00:14:54,995 you need trust, 398 00:14:55,870 --> 00:14:56,929 you need adoption, 399 00:14:57,309 --> 00:14:59,790 and you need tech that works. And you 400 00:14:59,790 --> 00:15:02,590 only know if your agentic system works and 401 00:15:02,590 --> 00:15:04,750 is actually solving the problem of the workflow 402 00:15:04,750 --> 00:15:05,570 and the practitioners 403 00:15:06,110 --> 00:15:08,534 if those practitioners are using it and are 404 00:15:08,534 --> 00:15:10,454 telling you what is working and what is 405 00:15:10,454 --> 00:15:12,615 not working, which requires a lot of trust 406 00:15:12,615 --> 00:15:14,954 and it requires a lot of adoption. So 407 00:15:15,254 --> 00:15:17,654 just as much investment in tech, I think, 408 00:15:17,654 --> 00:15:21,014 also needs investment in just change management with 409 00:15:21,014 --> 00:15:21,674 the organization. 410 00:15:22,529 --> 00:15:25,329 I'm very excited about our future here. I 411 00:15:25,329 --> 00:15:27,889 I think there is real opportunity here to 412 00:15:27,889 --> 00:15:31,490 personalize care and outreach and and help lots 413 00:15:31,490 --> 00:15:33,570 of people in the ways they like to 414 00:15:33,570 --> 00:15:34,149 be helped. 415 00:15:35,504 --> 00:15:36,164 Yeah. No. 416 00:15:36,544 --> 00:15:38,384 I can hear the excitement in your voice, 417 00:15:38,384 --> 00:15:40,625 Me too, and it's always it's a pleasure 418 00:15:40,625 --> 00:15:42,225 to talk to people who are really passionate 419 00:15:42,225 --> 00:15:44,384 about the work that they're doing. And and, 420 00:15:44,705 --> 00:15:47,200 I really appreciate also those last three pillars 421 00:15:47,200 --> 00:15:48,639 you shared with our listeners. You know, what 422 00:15:48,639 --> 00:15:50,639 will it take looking ahead? It'll take trust, 423 00:15:50,639 --> 00:15:52,100 adoption, and tech that works. 424 00:15:52,480 --> 00:15:54,559 Such a fantastic discussion. Thank you so much 425 00:15:54,559 --> 00:15:56,399 again, Mitu, for making the time for Becker's 426 00:15:56,399 --> 00:15:56,899 today. 427 00:15:57,279 --> 00:15:59,039 Thank you very much, and thank you very 428 00:15:59,039 --> 00:16:01,200 much for the opportunity. Really excited to be 429 00:16:01,200 --> 00:16:01,700 here. 430 00:16:02,315 --> 00:16:04,554 And we also wanna thank our podcast sponsor 431 00:16:04,554 --> 00:16:06,174 for today, Lilac Software. 432 00:16:06,715 --> 00:16:08,715 You can tune into more podcasts from Becker's 433 00:16:08,715 --> 00:16:13,215 Healthcare by visiting our podcast page at beckershospitalreview.com.