1 00:00:00,160 --> 00:00:02,720 Hi, everyone. This is Lucas Voss with Becker's 2 00:00:02,720 --> 00:00:04,480 Healthcare. Thanks so much for tuning in to 3 00:00:04,480 --> 00:00:07,139 the Becker's Healthcare podcast series. It's fantastic 4 00:00:07,440 --> 00:00:09,679 to have you. Today, we're talking about why 5 00:00:09,679 --> 00:00:12,480 Ambient AI took off in 2025, 6 00:00:12,480 --> 00:00:14,339 as we all know, and what's next 7 00:00:15,199 --> 00:00:16,695 in 2026 8 00:00:16,695 --> 00:00:18,774 and beyond. And joining me for today's discussion, 9 00:00:18,774 --> 00:00:20,214 I'm so excited to have him, is doctor 10 00:00:20,214 --> 00:00:22,935 Thomas Kelly, CEO and cofounder of Heidi. Doctor 11 00:00:22,935 --> 00:00:24,695 Kelly, thanks so much for taking some time 12 00:00:24,695 --> 00:00:26,054 and being here today. It's great to have 13 00:00:26,054 --> 00:00:26,554 you. 14 00:00:27,015 --> 00:00:28,935 Oh, it is. Yeah. No worries. Excited to 15 00:00:28,935 --> 00:00:30,730 chat more. For those that might not know 16 00:00:30,730 --> 00:00:32,409 you yet, I do wanna kick us off 17 00:00:32,409 --> 00:00:34,090 with introductions here. If you just wanna give 18 00:00:34,090 --> 00:00:37,229 us a brief overview of, yourself and certainly 19 00:00:37,289 --> 00:00:38,510 your work in health care. 20 00:00:39,210 --> 00:00:41,369 Yeah. So I'm Tom. I'm a doctor by 21 00:00:41,369 --> 00:00:41,869 background. 22 00:00:42,490 --> 00:00:44,250 Australian born, if you can't tell from the 23 00:00:44,250 --> 00:00:44,750 accent. 24 00:00:45,395 --> 00:00:48,114 Founded Heidi after being a vascular surgery resident 25 00:00:48,114 --> 00:00:49,494 in 2021. 26 00:00:49,634 --> 00:00:52,034 Had a view that every doctor should practice 27 00:00:52,034 --> 00:00:53,875 with some sort of sort of AI partner 28 00:00:53,875 --> 00:00:56,034 in their care. I had sci fi visions 29 00:00:56,034 --> 00:00:58,375 of, like, you know, maybe kinda everything, like 30 00:00:58,869 --> 00:01:02,090 prescribing, ordering investigations for me, reviewing my results, 31 00:01:02,629 --> 00:01:05,430 doing my notes, basically freeing me from my 32 00:01:05,430 --> 00:01:07,670 medical record and letting me see more patients, 33 00:01:07,670 --> 00:01:09,670 particularly for me, operate more because that's what 34 00:01:09,670 --> 00:01:12,709 I love doing. And, yeah, I'm proud to 35 00:01:12,709 --> 00:01:13,209 lead, 36 00:01:13,784 --> 00:01:15,465 probably one of, like, the largest health care 37 00:01:15,465 --> 00:01:17,625 companies in the world. So we we're a 38 00:01:17,625 --> 00:01:18,525 global company, 39 00:01:19,305 --> 00:01:22,504 around more than 50% of doctors in across 40 00:01:22,504 --> 00:01:25,465 Australia, New Zealand, Canada, UK use Heidi every 41 00:01:25,465 --> 00:01:27,459 week. And then in The US, we have 42 00:01:27,459 --> 00:01:29,000 an amazing growing business. 43 00:01:30,019 --> 00:01:31,859 I'd say where we're most prominent is in 44 00:01:31,859 --> 00:01:34,040 outpatient specialties, so across, like, urology, 45 00:01:34,739 --> 00:01:35,799 like, GI Alliance, 46 00:01:36,180 --> 00:01:37,000 US Urology, 47 00:01:37,540 --> 00:01:39,584 a a bunch of other amazing partners, and 48 00:01:39,584 --> 00:01:42,704 then, a few large health systems and growing, 49 00:01:42,864 --> 00:01:45,125 every year. So like Beth Israel, Franciscan, 50 00:01:45,424 --> 00:01:46,144 Maine General, 51 00:01:47,584 --> 00:01:49,025 Well, it's great to have you, and I 52 00:01:49,025 --> 00:01:51,344 feel like, again, your sci fi vision that 53 00:01:51,344 --> 00:01:53,659 you just mentioned sort of became true in 54 00:01:53,659 --> 00:01:54,619 2025 55 00:01:54,619 --> 00:01:56,619 for a lot of folks, for a lot 56 00:01:56,619 --> 00:01:58,159 of providers in itself. 57 00:01:58,700 --> 00:02:01,259 It's moved incredibly fast. It's going to move 58 00:02:01,259 --> 00:02:03,340 incredibly fast in 2026 59 00:02:03,340 --> 00:02:04,560 and certainly beyond. 60 00:02:05,084 --> 00:02:07,885 From your perspective and your work in the 61 00:02:07,885 --> 00:02:08,385 space, 62 00:02:09,245 --> 00:02:12,365 what's driving that pace right now, and what 63 00:02:12,365 --> 00:02:13,185 makes ambient 64 00:02:13,724 --> 00:02:14,384 AI fundamentally 65 00:02:14,685 --> 00:02:15,185 different 66 00:02:15,485 --> 00:02:17,665 sort of from what we've seen in earlier 67 00:02:17,724 --> 00:02:19,745 health care health care tech waves? 68 00:02:20,700 --> 00:02:22,319 Yeah. I think overwhelmingly, 69 00:02:23,099 --> 00:02:24,400 what I'm seeing is 70 00:02:24,780 --> 00:02:26,460 that, you know, and it's like in every 71 00:02:26,460 --> 00:02:28,780 industry, not just health care. So the boards 72 00:02:28,780 --> 00:02:30,860 and CEOs are asking their teams, like, what 73 00:02:30,860 --> 00:02:32,379 are we doing about AI? Like, I'm here 74 00:02:32,460 --> 00:02:34,485 I'm using it. It's amazing. Like, it's it's 75 00:02:34,485 --> 00:02:36,485 saving me time. Like, what what's your AI 76 00:02:36,485 --> 00:02:36,985 strategy? 77 00:02:37,364 --> 00:02:38,424 And almost always, 78 00:02:38,724 --> 00:02:40,965 what I've seen is that the especially in 79 00:02:40,965 --> 00:02:42,664 larger organizations, the CIOs, 80 00:02:43,405 --> 00:02:44,485 I love them, but they don't have a 81 00:02:44,485 --> 00:02:46,004 lot of taste. So they just end up 82 00:02:46,004 --> 00:02:48,405 picking whatever is the most integrated option in 83 00:02:48,405 --> 00:02:49,689 their stack, and then they go back to 84 00:02:49,689 --> 00:02:51,450 their board and they say, hey. Look. We 85 00:02:51,530 --> 00:02:54,010 we're doing AI now, and we've chosen chosen 86 00:02:54,010 --> 00:02:55,069 an Ambient tool. 87 00:02:55,370 --> 00:02:57,930 And specifically with Ambient, the reason I think 88 00:02:57,930 --> 00:03:00,170 it's gone so fast is it's one of 89 00:03:00,170 --> 00:03:03,164 the few things that doctors actually love. So 90 00:03:03,164 --> 00:03:05,104 usually usually software implementation 91 00:03:05,485 --> 00:03:07,564 is like a doctor's worst nightmare, like sitting 92 00:03:07,564 --> 00:03:09,245 in classes with someone teaching them how to 93 00:03:09,245 --> 00:03:11,324 click click different places in the medical record. 94 00:03:11,324 --> 00:03:12,844 It's not not why I did the went 95 00:03:12,844 --> 00:03:13,745 into the job. 96 00:03:14,044 --> 00:03:14,544 And, 97 00:03:14,844 --> 00:03:18,409 basically, I think that that downwards pressure met 98 00:03:18,550 --> 00:03:21,670 doctors wanting this tool, and so the CIOs 99 00:03:21,670 --> 00:03:24,069 generally picked whichever scribe work best with that 100 00:03:24,069 --> 00:03:26,710 medical record. I think this year and probably 101 00:03:26,710 --> 00:03:28,389 next year is gonna be where the rubber 102 00:03:28,389 --> 00:03:30,469 meets the road on ROI. Like, how many 103 00:03:30,469 --> 00:03:32,389 of our my doctors are actually using this? 104 00:03:32,389 --> 00:03:32,685 Like, 105 00:03:33,245 --> 00:03:35,245 are there better tools? Like, how should I 106 00:03:35,245 --> 00:03:37,965 actually deeply think about, like, what AI means 107 00:03:37,965 --> 00:03:39,905 for the organization, especially as 108 00:03:40,284 --> 00:03:40,944 the capabilities 109 00:03:41,245 --> 00:03:41,745 grow? 110 00:03:42,125 --> 00:03:42,784 I think 111 00:03:43,324 --> 00:03:45,004 back to my sci fi vision, I think 112 00:03:45,004 --> 00:03:47,889 it'll become more and more about, like, an 113 00:03:47,889 --> 00:03:50,610 AI partner that actually participates in care. Like, 114 00:03:50,610 --> 00:03:53,270 we're already seeing in Utah, like, the Doctronic 115 00:03:53,409 --> 00:03:56,150 trial where where an AI is starting to 116 00:03:56,370 --> 00:03:57,349 basically, like, 117 00:03:57,650 --> 00:04:00,689 almost fully autonomously prescribe with some doctor sign 118 00:04:00,689 --> 00:04:02,634 off. And so I think the health systems, 119 00:04:02,634 --> 00:04:04,875 like, that's how I would conceptually try to 120 00:04:04,875 --> 00:04:06,875 plan how I work with AI. Like, where 121 00:04:06,875 --> 00:04:08,394 are the areas in the health system that 122 00:04:08,394 --> 00:04:08,894 are 123 00:04:09,194 --> 00:04:12,074 most costly or or hardest to staff or 124 00:04:12,074 --> 00:04:13,514 have the longest wait lists, 125 00:04:14,419 --> 00:04:17,060 and spend time with with companies like ours 126 00:04:17,060 --> 00:04:19,300 to try to figure out, okay, like you 127 00:04:19,300 --> 00:04:20,660 know, because we already have it. Like, my 128 00:04:20,660 --> 00:04:22,579 cardi can read the record, can look up 129 00:04:22,579 --> 00:04:24,740 evidence, can call patients, can so it's like 130 00:04:24,740 --> 00:04:26,979 there's so many things you can do. It 131 00:04:26,979 --> 00:04:29,865 just depends on what your objectives are. 132 00:04:30,165 --> 00:04:32,725 And, yeah, it'll be interesting to see. Like, 133 00:04:32,725 --> 00:04:33,865 I do think that 134 00:04:34,245 --> 00:04:36,504 your play scribing on its own 135 00:04:37,044 --> 00:04:40,084 will struggle to find durable value because it's 136 00:04:40,084 --> 00:04:41,785 it's getting easier and easier 137 00:04:42,680 --> 00:04:44,839 to generate amazing notes. It used to be 138 00:04:44,839 --> 00:04:46,759 a really hard problem. Like, I spent two 139 00:04:46,759 --> 00:04:48,199 years of my life trying to make this 140 00:04:48,199 --> 00:04:48,699 work. 141 00:04:49,000 --> 00:04:50,680 Mhmm. But now if you if you just 142 00:04:50,680 --> 00:04:51,740 pick up an amazing 143 00:04:52,279 --> 00:04:53,259 foundation model, 144 00:04:53,959 --> 00:04:55,399 you know, you get most of the way 145 00:04:55,399 --> 00:04:55,899 there. 146 00:04:56,785 --> 00:04:58,704 So it's gotta be about the depth of 147 00:04:58,704 --> 00:05:00,384 the product, I think, and how much providers 148 00:05:00,384 --> 00:05:01,125 love it. 149 00:05:01,504 --> 00:05:03,425 So one of the things that was happening 150 00:05:03,425 --> 00:05:04,625 in 2025 151 00:05:04,625 --> 00:05:06,964 specifically was the fact that a lot of 152 00:05:07,105 --> 00:05:09,504 organizations were moving sort of from the pilot 153 00:05:09,504 --> 00:05:10,324 phase into 154 00:05:10,930 --> 00:05:14,129 adoption, into trying to get this at to 155 00:05:14,129 --> 00:05:15,509 at scale, essentially. 156 00:05:16,449 --> 00:05:19,089 Yeah. What's what's needed from leaders to make 157 00:05:19,089 --> 00:05:21,169 that happen? What are the conversations that need 158 00:05:21,169 --> 00:05:22,689 to happen to be able to say, okay. 159 00:05:22,689 --> 00:05:24,370 We're moving from a pilot project that was 160 00:05:24,370 --> 00:05:26,085 sort of a nice to have to then 161 00:05:26,085 --> 00:05:28,645 now, okay. We're using this daily, weekly, as 162 00:05:28,645 --> 00:05:30,025 you mentioned with some of the organizations 163 00:05:30,404 --> 00:05:31,464 that you work with. 164 00:05:32,324 --> 00:05:34,165 Yeah. I think we you definitely have to 165 00:05:34,165 --> 00:05:35,845 spend the time to nail it in in 166 00:05:35,845 --> 00:05:37,620 initial kind of so we we do, like, 167 00:05:37,620 --> 00:05:38,519 a stage rollout, 168 00:05:38,979 --> 00:05:39,800 where you have 169 00:05:40,180 --> 00:05:43,060 an initial alpha group, which is friend like, 170 00:05:43,060 --> 00:05:45,240 friendlies, so people in each of the departments. 171 00:05:46,180 --> 00:05:47,860 I think the key is not to cherry 172 00:05:47,860 --> 00:05:50,339 pick specialties. You have to have absolutely every 173 00:05:50,579 --> 00:05:53,324 everyone represented that you want to ultimately use 174 00:05:53,324 --> 00:05:53,904 the tool. 175 00:05:54,365 --> 00:05:56,064 Then in that phase, 176 00:05:56,365 --> 00:05:58,625 doing, like, real kind of AI enablement, 177 00:05:59,165 --> 00:06:01,245 setup for each of those specialists. So we 178 00:06:01,245 --> 00:06:03,404 do that with our CS and implementation team. 179 00:06:03,404 --> 00:06:05,725 So that's where we go to hematology and 180 00:06:05,725 --> 00:06:07,185 we say, oh, okay. For hematology, 181 00:06:07,965 --> 00:06:08,625 the visit 182 00:06:09,080 --> 00:06:10,620 contains very little, 183 00:06:11,000 --> 00:06:14,279 useful information. Actually, the value of Heidi would 184 00:06:14,279 --> 00:06:15,339 be if we 185 00:06:15,639 --> 00:06:17,639 look at all of the historical blood results 186 00:06:17,639 --> 00:06:19,800 and summarize them in the note. Like, that's 187 00:06:19,800 --> 00:06:22,074 a valuable thing. And then so then we 188 00:06:22,074 --> 00:06:25,055 set up materials and templates that teach hematologists 189 00:06:25,354 --> 00:06:26,954 how to use HYDIA in that way. So 190 00:06:26,954 --> 00:06:29,115 you can use the context tab and select 191 00:06:29,115 --> 00:06:31,594 the past sessions and then use this template, 192 00:06:31,594 --> 00:06:33,435 and it will summarize all the blood results 193 00:06:33,435 --> 00:06:35,459 just the way you like. And I think, 194 00:06:35,459 --> 00:06:37,379 to be honest, I think that's where we 195 00:06:37,379 --> 00:06:39,539 stand out. So if if you're working with 196 00:06:39,539 --> 00:06:41,379 a partner, it's like, oh, incredibly easy to 197 00:06:41,379 --> 00:06:43,220 roll out. Like, time to value is four 198 00:06:43,220 --> 00:06:45,459 weeks. You just turn it on, and everyone's 199 00:06:45,459 --> 00:06:46,979 gonna be using it. I feel like it's 200 00:06:46,979 --> 00:06:49,384 a red flag. Like, it's it's likely 201 00:06:49,925 --> 00:06:51,925 that they're just gonna be rolling out, like, 202 00:06:51,925 --> 00:06:54,504 the same template and, like, limited personalization 203 00:06:55,045 --> 00:06:56,665 for every single specialty, 204 00:06:57,204 --> 00:06:59,845 which means that when you hit a, I 205 00:06:59,845 --> 00:07:01,225 don't know, a plastic surgeon, 206 00:07:01,764 --> 00:07:05,230 like, the SOAP note isn't is not really 207 00:07:05,230 --> 00:07:07,069 that useful to them. So if you're just 208 00:07:07,069 --> 00:07:09,069 trying to drive them to use, like, a 209 00:07:09,069 --> 00:07:11,790 baked in, SOAP note generator that does a 210 00:07:11,790 --> 00:07:13,810 good problems list and captures better, 211 00:07:14,670 --> 00:07:16,830 problems from the record in order to generate 212 00:07:16,830 --> 00:07:18,625 more revenue, They're probably just not gonna use 213 00:07:18,625 --> 00:07:20,625 it because they hate the notes, and they 214 00:07:20,625 --> 00:07:22,785 they don't actually get it it doesn't write 215 00:07:22,785 --> 00:07:25,745 like them. It doesn't contain it's like more 216 00:07:25,745 --> 00:07:27,504 note bloat. Like, it's it's a lot of, 217 00:07:27,504 --> 00:07:30,305 like, random stuff, like random issues and problems 218 00:07:30,305 --> 00:07:32,004 that they they're not actually managing 219 00:07:32,389 --> 00:07:33,990 just because that helps with, like, the RCM 220 00:07:33,990 --> 00:07:35,990 down the track. It's like, I'm just dealing 221 00:07:35,990 --> 00:07:38,310 with this right hand fracture. Like, get out 222 00:07:38,310 --> 00:07:39,129 of my hair. 223 00:07:39,669 --> 00:07:41,930 So I think I think it depends on 224 00:07:42,149 --> 00:07:45,189 every from our perspective, every specialty is actually 225 00:07:45,189 --> 00:07:47,290 quite different. Need to treat them differently. 226 00:07:47,745 --> 00:07:49,904 That's how you get so we we generally 227 00:07:49,904 --> 00:07:51,524 see, like, 80% activation 228 00:07:51,824 --> 00:07:53,605 from from sign up. So if someone 229 00:07:54,225 --> 00:07:56,225 if if you promote it internally, you set 230 00:07:56,225 --> 00:07:58,705 it free. If someone signs up, roughly four 231 00:07:58,705 --> 00:08:00,165 out of five will use it, 232 00:08:00,625 --> 00:08:03,319 every week, and something like 70% of them 233 00:08:03,319 --> 00:08:04,379 will use it every day. 234 00:08:04,759 --> 00:08:06,839 And the the thing that's hard is, I 235 00:08:06,839 --> 00:08:07,339 think, 236 00:08:07,879 --> 00:08:10,279 in the the PR and the marketing, it's 237 00:08:10,279 --> 00:08:12,520 sometimes hard to trust because if I talk 238 00:08:12,520 --> 00:08:13,960 to the systems and they're like, oh, yeah. 239 00:08:13,960 --> 00:08:16,154 It's only in primary care, but then they'll 240 00:08:16,154 --> 00:08:17,835 then they'll promote it as if it's, like, 241 00:08:17,835 --> 00:08:20,714 80% adoption in every every market. Yeah. Primary 242 00:08:20,714 --> 00:08:22,975 care, easy. Like, they they see everything. 243 00:08:23,355 --> 00:08:25,535 They do high, turnover medicine. 244 00:08:25,915 --> 00:08:28,235 This standard soap note is amazing for them. 245 00:08:28,235 --> 00:08:29,595 But if you think about, like, all the 246 00:08:29,595 --> 00:08:30,095 specialties, 247 00:08:30,839 --> 00:08:33,000 all of the different use cases, you have 248 00:08:33,000 --> 00:08:35,660 to have some ability to personalize, generate templates 249 00:08:35,799 --> 00:08:36,299 that 250 00:08:36,919 --> 00:08:39,000 live outside of the medical record. They still 251 00:08:39,000 --> 00:08:41,160 use data from the medical record, but there 252 00:08:41,160 --> 00:08:43,660 are many things that providers do that 253 00:08:44,134 --> 00:08:46,294 just require, like, a more flexible AI tool. 254 00:08:46,294 --> 00:08:48,554 So that's usually where we sit in. So 255 00:08:48,855 --> 00:08:50,934 if if a system is is churning off 256 00:08:50,934 --> 00:08:52,774 another tool or they haven't picked one yet 257 00:08:52,774 --> 00:08:54,634 and they have a lot of specialty practice 258 00:08:55,174 --> 00:08:57,034 or they they've tried 259 00:08:57,414 --> 00:09:00,440 the, you know, single template best at RCM 260 00:09:00,660 --> 00:09:02,660 tool and they haven't actually seen much, 261 00:09:03,300 --> 00:09:06,820 much adoption or much uplift, then then, yeah, 262 00:09:07,059 --> 00:09:09,000 I think we're a good option. 263 00:09:09,460 --> 00:09:09,960 Yeah. 264 00:09:10,660 --> 00:09:12,274 I wanna come back to the adoption piece 265 00:09:12,274 --> 00:09:13,955 and the and the usability part of it, 266 00:09:13,955 --> 00:09:15,875 which you've mentioned is is really crucial. Right? 267 00:09:15,875 --> 00:09:18,115 It needs to have that aspect to it 268 00:09:18,115 --> 00:09:19,875 for for it to work on a on 269 00:09:19,875 --> 00:09:21,014 a scale basis. 270 00:09:21,955 --> 00:09:24,995 For you personally, what does usable mean in 271 00:09:24,995 --> 00:09:27,330 day to day practice? And then certainly, we've 272 00:09:27,330 --> 00:09:29,429 talked about the workflow piece of this. 273 00:09:29,730 --> 00:09:32,870 How should leaders think about workflow fit versus 274 00:09:33,009 --> 00:09:35,029 then also the EHR integration 275 00:09:35,809 --> 00:09:37,269 when they're looking at evaluating 276 00:09:37,809 --> 00:09:38,950 ambient AI? 277 00:09:39,875 --> 00:09:42,034 Yeah. Yeah. I think for me so having 278 00:09:42,034 --> 00:09:42,855 been a doctor, 279 00:09:43,475 --> 00:09:46,195 the main thing we time we save is 280 00:09:46,195 --> 00:09:46,695 when 281 00:09:47,075 --> 00:09:48,214 the amount of time 282 00:09:48,514 --> 00:09:50,934 writing notes. So, basically, like, keyboard strokes, 283 00:09:51,315 --> 00:09:52,980 how many times do I have to edit 284 00:09:53,059 --> 00:09:53,559 change, 285 00:09:53,940 --> 00:09:56,179 how much of what the system writes am 286 00:09:56,179 --> 00:09:57,240 I actually accepting, 287 00:09:57,779 --> 00:09:59,299 and, of course, like, how often am I 288 00:09:59,299 --> 00:10:00,899 using it? Like, is it am I only 289 00:10:00,899 --> 00:10:02,659 cherry picking it for certain visits? Am I 290 00:10:02,659 --> 00:10:03,959 using it for every visit? 291 00:10:04,259 --> 00:10:04,759 And 292 00:10:05,154 --> 00:10:06,995 we we measure that on, like, what's our 293 00:10:06,995 --> 00:10:09,735 daily active usage? Like, what percentage of sessions 294 00:10:09,955 --> 00:10:12,355 are we capturing? And and we should be 295 00:10:12,355 --> 00:10:15,394 in most visits unless there's, you know, patient 296 00:10:15,394 --> 00:10:17,554 who hasn't consented or it's a really tricky 297 00:10:17,554 --> 00:10:18,615 tricky circumstance. 298 00:10:19,394 --> 00:10:19,894 So 299 00:10:20,420 --> 00:10:22,899 so for us, that has always meant that 300 00:10:22,899 --> 00:10:25,139 we don't really bend the knee and, like, 301 00:10:25,139 --> 00:10:28,259 put Heidi so integrated into the medical record 302 00:10:28,259 --> 00:10:29,879 that it is, like, unusably 303 00:10:30,340 --> 00:10:32,200 rigid. So, like, 304 00:10:32,834 --> 00:10:34,754 without naming names, I would say most of 305 00:10:34,754 --> 00:10:37,235 the options that are very, very, very tightly 306 00:10:37,235 --> 00:10:39,334 integrated with the record, like the default options, 307 00:10:39,794 --> 00:10:43,074 the product teams of those products, they live 308 00:10:43,074 --> 00:10:45,154 and die by the medical record. So they 309 00:10:45,154 --> 00:10:46,740 wake up in the morning and they think, 310 00:10:46,980 --> 00:10:48,820 if this tool is doing its job, it 311 00:10:48,820 --> 00:10:51,139 just fills the fields in front of the 312 00:10:51,139 --> 00:10:53,240 doctor. Like, that's how they view the world. 313 00:10:53,620 --> 00:10:55,639 Now don't get me wrong. That is valuable 314 00:10:55,700 --> 00:10:58,740 for sure. But I view the world more 315 00:10:58,740 --> 00:11:01,480 in the, like, consumer AI lens like Claude 316 00:11:01,485 --> 00:11:04,464 or Chativity or, actually, I think I view, 317 00:11:04,605 --> 00:11:06,365 like, some of the more direct to doctor 318 00:11:06,365 --> 00:11:09,804 competitors as as real competitors for us. So 319 00:11:09,804 --> 00:11:10,704 it's like, okay. 320 00:11:11,164 --> 00:11:14,464 Ignoring, like, the, CDI and the RCM 321 00:11:14,820 --> 00:11:16,580 flow that happens and all the reasons we 322 00:11:16,580 --> 00:11:18,679 have our medic record, in ten years, 323 00:11:19,139 --> 00:11:21,220 is a doctor actually gonna be sitting on 324 00:11:21,220 --> 00:11:23,059 a screen that has all these fields that 325 00:11:23,059 --> 00:11:25,460 require to be filled? Like, definitely not. I 326 00:11:25,460 --> 00:11:27,399 have failed if that is still the case. 327 00:11:27,940 --> 00:11:31,355 So my take is that that you have 328 00:11:31,355 --> 00:11:33,355 to build, like, a different UI. Like, for 329 00:11:33,355 --> 00:11:35,054 someone who hasn't hasn't, 330 00:11:35,674 --> 00:11:38,334 done engineering, there's this application called Cursor, 331 00:11:38,714 --> 00:11:40,875 which is what a lot of like, almost 332 00:11:41,115 --> 00:11:42,954 I don't know. Most engineers are now using 333 00:11:42,954 --> 00:11:45,190 this tool. And, basically, it's like a single 334 00:11:45,190 --> 00:11:47,110 chat field, and you can turn on voice. 335 00:11:47,110 --> 00:11:49,350 And you you speak to the system, you 336 00:11:49,350 --> 00:11:51,029 type what you want it to do, and 337 00:11:51,029 --> 00:11:52,649 it's fully abstracting away 338 00:11:53,269 --> 00:11:55,190 something like five to 10 different pieces of 339 00:11:55,190 --> 00:11:57,004 software for you. And they all look like 340 00:11:57,004 --> 00:11:58,684 databases. Like, they all look like the medical 341 00:11:58,684 --> 00:12:00,764 record. They have all these different things that 342 00:12:00,764 --> 00:12:02,605 you would have historically had to tick and 343 00:12:02,605 --> 00:12:04,445 click and write in this file and write 344 00:12:04,445 --> 00:12:06,445 another file and then so it's a very 345 00:12:06,445 --> 00:12:08,764 similar idea. And so Heidi is the same. 346 00:12:08,764 --> 00:12:11,519 Like, usability means that we build AI first. 347 00:12:11,519 --> 00:12:14,000 Like, what what not just MD and AI, 348 00:12:14,000 --> 00:12:15,360 but, like, what are all the things that 349 00:12:15,360 --> 00:12:18,000 will become possible with this tool? Yes. We 350 00:12:18,000 --> 00:12:20,000 have to integrate very deeply. Like, we have 351 00:12:20,000 --> 00:12:21,120 to be able to order. We have to 352 00:12:21,120 --> 00:12:22,559 be able to do problems list. We have 353 00:12:22,559 --> 00:12:24,480 to be able to write notes that generate 354 00:12:24,480 --> 00:12:26,865 more revenue, capture issues that we would have 355 00:12:26,865 --> 00:12:27,404 otherwise missed, 356 00:12:28,105 --> 00:12:30,985 all things we do. But but, fundamentally, like, 357 00:12:30,985 --> 00:12:33,625 we will not sacrifice making Heidi, like, some, 358 00:12:33,625 --> 00:12:34,445 like, ribbon 359 00:12:34,985 --> 00:12:37,144 inside of the medical record that does AI. 360 00:12:37,144 --> 00:12:39,144 Like, that's just not how it's gonna happen. 361 00:12:39,144 --> 00:12:40,605 It's gonna be a full screen 362 00:12:40,919 --> 00:12:43,080 experience that is, like, built for AI, that's 363 00:12:43,080 --> 00:12:43,580 fast, 364 00:12:44,039 --> 00:12:46,360 that does, like, agent calls really nicely, that 365 00:12:46,360 --> 00:12:48,700 shows you the thinking, that does, like, sourcing, 366 00:12:48,839 --> 00:12:51,559 that it's all, like, one thing, and that's 367 00:12:51,559 --> 00:12:52,779 the kind of, like, superintelligence 368 00:12:53,399 --> 00:12:55,875 AI care partner that I imagine every doctor 369 00:12:55,875 --> 00:12:56,774 brings to work. 370 00:12:57,235 --> 00:12:59,715 And so when when evaluating I know this 371 00:12:59,715 --> 00:13:02,675 sounds like evangelical and, you know, I'm I'm, 372 00:13:02,995 --> 00:13:05,394 I'm pitching on where it goes. I do 373 00:13:05,394 --> 00:13:07,679 think it matters. So the the things I 374 00:13:07,679 --> 00:13:10,579 would watch out for is if it's incredibly 375 00:13:10,799 --> 00:13:12,639 integrated in the medical record, then you should 376 00:13:12,639 --> 00:13:14,159 just buy the one in the medical record. 377 00:13:14,159 --> 00:13:15,539 Like, that's gonna exist, 378 00:13:15,919 --> 00:13:17,839 you know, whether it's Epic's or not, whoever. 379 00:13:17,839 --> 00:13:19,679 Like, they're gonna do that the best. I 380 00:13:19,679 --> 00:13:22,254 don't really think the companies whose value prop 381 00:13:22,254 --> 00:13:24,575 is there more integrated with Epic than Epic 382 00:13:24,575 --> 00:13:27,235 Zone AI Scribe are gonna be durable, really, 383 00:13:27,534 --> 00:13:28,914 unless they do other things. 384 00:13:29,455 --> 00:13:31,455 If that's failed, so if you tried that 385 00:13:31,455 --> 00:13:33,455 or or your doctors say, I don't really 386 00:13:33,455 --> 00:13:35,529 like like this approach. Like, I wanna be 387 00:13:35,529 --> 00:13:37,129 able to ask AI question. I wanna be 388 00:13:37,129 --> 00:13:38,330 able to look something up. I wanna be 389 00:13:38,330 --> 00:13:40,009 able to summarize my week. I wanna be 390 00:13:40,009 --> 00:13:40,750 able to, 391 00:13:41,529 --> 00:13:43,610 ask about the last three sessions and then, 392 00:13:43,610 --> 00:13:45,950 like, change the template because of this patient 393 00:13:46,090 --> 00:13:47,690 the last three times I spoke about the 394 00:13:47,690 --> 00:13:49,304 diabetes, but I don't want to summarize the 395 00:13:49,304 --> 00:13:52,345 whole year. Like, usability is about, like, I 396 00:13:52,345 --> 00:13:54,205 am an AI serving my doctor. 397 00:13:54,665 --> 00:13:57,225 And, yes, I ultimately send data back to 398 00:13:57,225 --> 00:13:59,465 the record, but I have a flexible interface 399 00:13:59,465 --> 00:14:00,745 that I can let the doctor do what 400 00:14:00,745 --> 00:14:02,764 they want and then ultimately send it back. 401 00:14:03,110 --> 00:14:04,470 So does that make sense? I don't know. 402 00:14:04,470 --> 00:14:06,149 I'm rambling a bit, but I think that's 403 00:14:06,149 --> 00:14:08,310 the that's the gist. That's your sci fi 404 00:14:08,310 --> 00:14:09,830 that's your sci fi vision. I think it 405 00:14:09,830 --> 00:14:11,750 does make sense. It's very important. It it 406 00:14:11,750 --> 00:14:13,910 creates that flexibility for users. It's not a 407 00:14:13,910 --> 00:14:16,550 rigid experience. It it it's it's tailored, and 408 00:14:16,550 --> 00:14:18,389 that's as you mentioned, I think that's the 409 00:14:18,389 --> 00:14:19,774 way it's it's going to go in the 410 00:14:19,774 --> 00:14:21,375 next ten years. It's not going to be 411 00:14:21,375 --> 00:14:23,454 rigid fields, as you've mentioned, that are going 412 00:14:23,454 --> 00:14:24,894 to be be filled out. It's a it's 413 00:14:24,894 --> 00:14:27,875 a flexible flexible process that needs to happen. 414 00:14:27,934 --> 00:14:30,414 Yeah. One interesting thing that you've mentioned, and 415 00:14:30,414 --> 00:14:32,514 I think this is very important, is 416 00:14:32,894 --> 00:14:33,714 that oftentimes 417 00:14:34,659 --> 00:14:38,019 these early wins, everything works, everything is great, 418 00:14:38,019 --> 00:14:40,519 can actually be a red flag for organizations 419 00:14:40,659 --> 00:14:42,580 in terms of, oh, this is everything's working. 420 00:14:42,580 --> 00:14:44,820 We have the early results. Yeah. We've seen 421 00:14:44,820 --> 00:14:45,960 we've seen, you know, 422 00:14:46,340 --> 00:14:48,740 after hours EHR time is is reduced, all 423 00:14:48,740 --> 00:14:51,764 of a sudden improved documentation, completeness, etcetera. It's 424 00:14:51,764 --> 00:14:52,904 all there. Right? 425 00:14:53,285 --> 00:14:55,845 When you're looking at results like these and 426 00:14:55,845 --> 00:14:57,445 sort of these when you think about the 427 00:14:57,445 --> 00:14:58,504 red flag, right, 428 00:14:58,884 --> 00:15:01,684 what should health care leaders ask right now 429 00:15:01,684 --> 00:15:04,509 to separate that, you know, early win sort 430 00:15:04,509 --> 00:15:05,649 of from repeatable 431 00:15:06,110 --> 00:15:08,450 durable impact for the next year? 432 00:15:09,309 --> 00:15:10,850 Yeah. I would ask so 433 00:15:11,470 --> 00:15:13,889 how many of the doctors across all specialties 434 00:15:14,029 --> 00:15:15,470 are using this tool, and how many of 435 00:15:15,470 --> 00:15:16,990 them like, what's what are the what's the 436 00:15:16,990 --> 00:15:19,709 actual denominator of the activation number? Like, is 437 00:15:19,709 --> 00:15:22,214 it everyone who's signed in, which I think 438 00:15:22,214 --> 00:15:24,455 is, like, the right because I, you know, 439 00:15:24,455 --> 00:15:25,975 I can only do so much to get 440 00:15:25,975 --> 00:15:27,735 someone to to actually sign up to the 441 00:15:27,735 --> 00:15:30,054 tool. But once they sign up, then then, 442 00:15:30,054 --> 00:15:31,894 yeah, like, if I'm activating most of the 443 00:15:31,894 --> 00:15:34,240 users, I've succeeded. So you'd wanna know that 444 00:15:34,240 --> 00:15:35,700 it's working across all specialties. 445 00:15:36,559 --> 00:15:38,100 You know, what's the, 446 00:15:38,480 --> 00:15:40,559 there's, like, UX, which is just, like, what's 447 00:15:40,559 --> 00:15:41,459 the user experience? 448 00:15:42,159 --> 00:15:43,600 Do they find it easy to use? Do 449 00:15:43,600 --> 00:15:46,019 they find it valuable? Would they be frustrated 450 00:15:46,079 --> 00:15:46,720 if if, 451 00:15:47,440 --> 00:15:48,419 it's taken away? 452 00:15:48,945 --> 00:15:50,644 You can look at objective measures. 453 00:15:51,184 --> 00:15:53,424 It's it's shocking to me seeing some of 454 00:15:53,424 --> 00:15:55,684 the studies that are coming out about, like, 455 00:15:56,065 --> 00:15:59,524 how little some some products are actually impacting 456 00:15:59,825 --> 00:16:01,845 both, like, revenue, time, and record. 457 00:16:02,940 --> 00:16:05,419 And they universally, doctors love it because they 458 00:16:05,419 --> 00:16:07,419 don't have to type as much, but, actually, 459 00:16:07,419 --> 00:16:10,139 that's not saving them as maybe it's not 460 00:16:10,139 --> 00:16:12,879 generating as much ROI as as maybe people 461 00:16:12,940 --> 00:16:15,019 thought months thought it would. The reason it's 462 00:16:15,019 --> 00:16:17,019 shocking for me is because I've only seen 463 00:16:17,019 --> 00:16:19,335 the studies that I've seen, and, again, they're, 464 00:16:19,335 --> 00:16:21,514 like, in that bucket of, like, very integrated 465 00:16:21,735 --> 00:16:23,654 world. And I think it makes sense because 466 00:16:23,654 --> 00:16:25,575 you still have to review every field. You 467 00:16:25,575 --> 00:16:27,495 have to, like, click all the drop downs. 468 00:16:27,495 --> 00:16:28,935 Like, you're stuck in the record. So you 469 00:16:28,935 --> 00:16:30,535 can only get so much of a speed 470 00:16:30,535 --> 00:16:31,759 up. Whereas in our, 471 00:16:32,300 --> 00:16:32,800 randomized, 472 00:16:34,060 --> 00:16:36,540 papers, like, in across the world, we see, 473 00:16:36,540 --> 00:16:37,279 like, overwhelmingly 474 00:16:37,899 --> 00:16:40,620 in improved, like, RBUs and time in record 475 00:16:40,620 --> 00:16:42,779 goes down massively. So that I think that's 476 00:16:42,779 --> 00:16:45,340 the story for 2026, 2027. That's it's up 477 00:16:45,340 --> 00:16:47,955 to me to tell it. But, basically, people 478 00:16:47,955 --> 00:16:49,095 who think that 479 00:16:49,794 --> 00:16:51,875 AI that is gonna change health care in 480 00:16:51,875 --> 00:16:53,875 the medical record is, like, a bad take, 481 00:16:53,875 --> 00:16:55,634 it's going to be that the records are 482 00:16:55,634 --> 00:16:56,695 gonna get increasingly 483 00:16:57,554 --> 00:17:00,514 almost, like, subjugated to the back and definitely 484 00:17:00,514 --> 00:17:01,975 used by admin and revenue, 485 00:17:02,429 --> 00:17:04,430 definitely exists. Like, they will always be there. 486 00:17:04,430 --> 00:17:06,430 Like, we will be great partners, but it's 487 00:17:06,430 --> 00:17:08,430 about this, like, AI layer that sits on 488 00:17:08,430 --> 00:17:10,829 top that partners with the doctor to mean 489 00:17:10,829 --> 00:17:12,269 that they do not need to use the 490 00:17:12,269 --> 00:17:13,890 record. Like, that is success. 491 00:17:14,474 --> 00:17:16,554 Success is I do not ever sit on 492 00:17:16,554 --> 00:17:18,654 a computer. Like, that's what a doctor wants. 493 00:17:18,954 --> 00:17:21,035 So you can't build that if you're you 494 00:17:21,035 --> 00:17:22,634 are the record. Like, that is like it's 495 00:17:22,634 --> 00:17:24,154 like you're you're trying to get rid of 496 00:17:24,154 --> 00:17:26,474 yourself. So I think I think that's why, 497 00:17:26,474 --> 00:17:28,714 like, companies like HYD will will exist is, 498 00:17:28,714 --> 00:17:29,934 like, to try to 499 00:17:30,450 --> 00:17:32,849 to actually, like, strengthen the record, like, put 500 00:17:32,849 --> 00:17:35,409 more data into it. Like, actually, like, increase 501 00:17:35,409 --> 00:17:37,490 its value to the org, but then to 502 00:17:37,490 --> 00:17:38,150 the doctors, 503 00:17:38,609 --> 00:17:40,690 breathe in to actually just practice and, like, 504 00:17:40,690 --> 00:17:41,589 get the satisfaction. 505 00:17:41,970 --> 00:17:43,730 So so, basically, I would say, look at 506 00:17:43,730 --> 00:17:45,134 those those objective metrics 507 00:17:45,615 --> 00:17:47,775 and then question, like, is there a good 508 00:17:47,775 --> 00:17:49,454 ROI case here? Like, am I getting the 509 00:17:49,454 --> 00:17:51,375 value out of the tool that maybe all 510 00:17:51,375 --> 00:17:53,295 the contracts maybe that I've committed to? And 511 00:17:53,295 --> 00:17:56,095 then just, yeah, raise raise your ambition and 512 00:17:56,095 --> 00:17:58,174 expectation for your partners and look at what's 513 00:17:58,174 --> 00:18:01,130 out there. Because I'd say for most cases, 514 00:18:01,269 --> 00:18:03,349 Heidi wouldn't have been one that they considered 515 00:18:03,349 --> 00:18:05,269 when their CEO said, hey. Choose an AI 516 00:18:05,269 --> 00:18:07,429 tool. At least in The US, everywhere else, 517 00:18:07,429 --> 00:18:08,950 yes, but in in The US, no, because 518 00:18:08,950 --> 00:18:10,970 we were just starting. And I think 519 00:18:11,349 --> 00:18:13,429 we have acquired a different perspective. So if 520 00:18:13,429 --> 00:18:14,649 you've had a bad time 521 00:18:14,974 --> 00:18:15,955 with a vendor, 522 00:18:16,335 --> 00:18:18,515 chances are you will have a bad time 523 00:18:18,654 --> 00:18:20,654 with the next vendor if the next one 524 00:18:20,654 --> 00:18:23,055 you choose is essentially the same thing. Like, 525 00:18:23,055 --> 00:18:25,535 we work super well with Epic. It's like 526 00:18:25,535 --> 00:18:27,535 that's not you already chose the one that 527 00:18:27,535 --> 00:18:29,055 works super well with Epic. Like, why is 528 00:18:29,055 --> 00:18:31,240 that gonna be any different? So I I 529 00:18:31,240 --> 00:18:32,599 think that's the way to think of it. 530 00:18:32,599 --> 00:18:33,099 Like, 531 00:18:33,640 --> 00:18:36,119 if if you're having success with the current 532 00:18:36,119 --> 00:18:38,599 version, then that's great. Like, I'm I'm happy. 533 00:18:38,839 --> 00:18:40,679 I just want doctors to have have good 534 00:18:40,679 --> 00:18:42,539 lives and and see more patients. So, 535 00:18:43,880 --> 00:18:45,880 yeah, I think just what is it? Always, 536 00:18:45,880 --> 00:18:46,944 like, measure 537 00:18:47,244 --> 00:18:50,125 think think about the metrics that matter to 538 00:18:50,125 --> 00:18:52,125 you and then talk to other partners and 539 00:18:52,125 --> 00:18:54,045 see see if there's a a better option 540 00:18:54,045 --> 00:18:54,625 out there. 541 00:18:55,005 --> 00:18:57,565 Be a little adventurous and buy into doctor 542 00:18:57,565 --> 00:18:59,484 Kelly's sci fi vision. I think that's the 543 00:18:59,484 --> 00:19:01,759 way to go. Yeah. Yeah. Doctor Kelly, thank 544 00:19:01,759 --> 00:19:02,960 you so much for your time. It's so 545 00:19:02,960 --> 00:19:04,559 fantastic to have you. It's a it's a 546 00:19:04,559 --> 00:19:06,339 great great conversation. Thank you. 547 00:19:06,880 --> 00:19:09,200 Thanks, Lucas. Appreciate it. And we also want 548 00:19:09,200 --> 00:19:10,960 to thank our podcast sponsor, Heidi. You can 549 00:19:10,960 --> 00:19:13,039 tune in to more podcasts from Becker's Healthcare 550 00:19:13,119 --> 00:19:16,979 by visiting our podcast page at beggershospitalreview.com.