WEBVTT - Philips North America's Jeff DiLullo Talks AI in Healthcare 0:00:02.480 --> 0:00:08.440 Bloomberg Audio Studios, podcasts, radio news. In the meantime, we 0:00:08.440 --> 0:00:12.000 want to continue with AI. Phillip's eleventh annual Future Health Index. 0:00:12.039 --> 0:00:14.680 It is out now. It examines how AI is already 0:00:14.720 --> 0:00:17.319 changing the way clinicians work, how they deliver care and 0:00:17.360 --> 0:00:20.920 manage pressure across the healthcare system. Now, among its findings, 0:00:21.079 --> 0:00:25.200 how healthcare executives are discovering that AI just isn't about 0:00:25.239 --> 0:00:28.440 cost savings. So let's get into it. Jeff Delulo is 0:00:28.480 --> 0:00:31.120 CEO at Phillips North America. It is the largest regional 0:00:31.240 --> 0:00:34.199 market and division of the Dutch health tech company Giant 0:00:34.200 --> 0:00:36.959 Phillips NV. Those ADRs trade in the US. They're done 0:00:36.960 --> 0:00:39.080 about three percent year to date. As we said, jeffs 0:00:39.120 --> 0:00:41.360 joins us in here in studio. Welcome, Welcome, How are 0:00:41.400 --> 0:00:42.440 you wonderful? 0:00:42.479 --> 0:00:43.159 Thanks for having me. 0:00:43.280 --> 0:00:45.280 Well, it's good to have you here. We do feel 0:00:45.320 --> 0:00:47.440 like we are spending so much time talking about AI. 0:00:47.479 --> 0:00:50.480 It's just coming into every aspect of our world. Tell 0:00:50.560 --> 0:00:52.960 us so about the study that you guys do. It 0:00:53.040 --> 0:00:55.640 is your eleventh annual. What it's all about, who you 0:00:55.680 --> 0:00:57.480 talk to and what are some of the findings. 0:00:57.800 --> 0:00:59.639 Well, thank you for having Carol. I want to set 0:00:59.680 --> 0:01:01.920 the table here very quickly, because what I talk to 0:01:01.960 --> 0:01:05.199 healthcare CEOs on a weekly basis, and the same three 0:01:05.240 --> 0:01:08.240 things always are putting pressure on that system. As you said, 0:01:08.880 --> 0:01:11.600 they're delivering more care longer to people that are living longer, 0:01:11.640 --> 0:01:14.080 and that complex complexity of that care is increasing, so 0:01:14.120 --> 0:01:17.279 that's actually putting stress on the system. They have staff 0:01:17.280 --> 0:01:20.080 shortages which continue to persist and get worse post pandemic, 0:01:20.480 --> 0:01:23.000 and then the cost of delivering that care is actually 0:01:23.080 --> 0:01:25.120 continuing to go up. So all those three things are 0:01:25.160 --> 0:01:27.360 really putting stress on the system. But the opportunity that 0:01:27.400 --> 0:01:30.880 we see is AI in healthcare is becoming quite of age. 0:01:30.880 --> 0:01:34.240 Where a year ago in this study, roughly twenty thousand people, 0:01:34.280 --> 0:01:39.080 two thousand clinicians, eighteen thousand patients, And what we saw 0:01:39.160 --> 0:01:41.480 from year to year is they've gone from experimenting and 0:01:41.520 --> 0:01:46.000 piloting and AI applications to add much more broad adoption 0:01:46.440 --> 0:01:49.840 and impact impact in time back they can spend with patients, 0:01:49.920 --> 0:01:54.600 impact in quality, diagnosis, and impact in well being, And 0:01:54.600 --> 0:01:56.320 they've been very clear about that in the study. 0:01:56.960 --> 0:02:00.480 What do those AI applications in healthcare actually look like 0:02:00.760 --> 0:02:02.080 in practice? 0:02:02.280 --> 0:02:05.840 So I'll give you a simple A simple example that 0:02:06.040 --> 0:02:09.679 everybody can relate to. So the average physician in this 0:02:09.760 --> 0:02:13.120 study said that they're getting five more patients per week 0:02:13.200 --> 0:02:15.240 because of tools that they're deploying in their practice. So 0:02:15.280 --> 0:02:18.240 when I go to my general practitioner, I get maybe 0:02:18.280 --> 0:02:20.560 ten or fifteen minutes if I'm lucky with him for 0:02:20.960 --> 0:02:23.120 and most of that time it's sitting on the keyboard 0:02:23.200 --> 0:02:25.840 asking me questions. Right, what we're seeing is they're getting 0:02:25.840 --> 0:02:28.960 the chance to ambient speak. What they're talking that dialogue, 0:02:28.960 --> 0:02:31.239 it's translating it into the medical record. So what they're 0:02:31.280 --> 0:02:33.440 doing is they're spending time looking you in the eye, 0:02:33.480 --> 0:02:37.120 asking you more questions. That's better for more patients they 0:02:37.120 --> 0:02:39.720 can see because they're spending quality time in the room, 0:02:40.200 --> 0:02:42.880 or maybe even more patients like we're seeing. 0:02:43.000 --> 0:02:46.320 Do doctors feel or medical professionals Jeff feel that we 0:02:46.400 --> 0:02:49.080 are just barely scratching the surface when it comes to 0:02:49.120 --> 0:02:50.079 AI in healthcare. 0:02:50.600 --> 0:02:53.519 Yeah, a prudent approach is good because you want to 0:02:53.520 --> 0:02:55.880 make sure that you have responsible AD deployment. And we're 0:02:56.040 --> 0:02:57.920 hugely as a tech company. We want to make sure 0:02:57.919 --> 0:03:00.160 it's responsible. What I actually think, we're the scraps at 0:03:00.200 --> 0:03:01.920 the table and the feast is yet to come. 0:03:02.000 --> 0:03:03.120 What do you think that feast looks like? 0:03:03.240 --> 0:03:05.760 So take the life of a radiologist. So the average 0:03:05.800 --> 0:03:09.639 radiologist has to work all day, sees patients, maybe do procedures. 0:03:09.880 --> 0:03:11.920 Then they take all these studies, two hundred studies home 0:03:11.919 --> 0:03:13.440 at night and they have to go these go through 0:03:13.440 --> 0:03:16.120 these studies. That's pajama time, right, it's there. It's long 0:03:16.200 --> 0:03:17.919 days for radiologists, which. 0:03:17.800 --> 0:03:20.120 It makes me always nervous as a radiologist who might 0:03:20.120 --> 0:03:22.320 be tired exactly and reading my like. 0:03:22.360 --> 0:03:25.760 I yeah, exactly, So you have the real days. They's stressful, 0:03:25.760 --> 0:03:26.040 and I know. 0:03:26.040 --> 0:03:28.280 There's great doctors out there, so I don't want to diss. 0:03:28.040 --> 0:03:30.520 This, but the system that's meant for them to have balance. 0:03:30.520 --> 0:03:32.600 And so what we're seeing is AI in the in 0:03:32.639 --> 0:03:34.680 the work, in the pathway, the whole. If you if 0:03:34.720 --> 0:03:36.200 you don't think of AI as a bunch of point 0:03:36.240 --> 0:03:39.280 solutions that do discrete things in healthcare, it has to 0:03:39.320 --> 0:03:42.440 fit to the radiologist. It has to fit into their workflow. 0:03:42.480 --> 0:03:45.200 So I can I can scan somebody with great AI 0:03:45.320 --> 0:03:47.240 much faster I get it into their workflow. I can 0:03:47.240 --> 0:03:49.480 determine who the most important person is to look at. 0:03:49.680 --> 0:03:52.000 I can find the slices in their scan that are 0:03:52.040 --> 0:03:54.360 the most important things to look at and even recommend. 0:03:54.600 --> 0:03:56.720 We think it looks like this, you should look right here. 0:03:56.920 --> 0:03:58.720 If you can do that at scale for two hundred 0:03:58.720 --> 0:04:02.320 studies a day, you're improving quality life that radiologist in 0:04:02.360 --> 0:04:03.960 the study, a third of them say they take a 0:04:03.960 --> 0:04:06.320 lot less work home and they're suffering less burnout and 0:04:06.360 --> 0:04:09.360 they're just charged in their in their balance. That also 0:04:09.400 --> 0:04:11.360 adds to quality exactly like you're saying. 0:04:11.920 --> 0:04:18.719 I also feel like the data that is input right, Like, 0:04:19.000 --> 0:04:22.600 there's just more data sets that scans can be compared against, 0:04:22.920 --> 0:04:25.080 whether it's an MRI, right, Like, do you know what 0:04:25.120 --> 0:04:25.560 I'm saying? 0:04:25.600 --> 0:04:29.840 That data coming you can use, but AI can use it, 0:04:30.040 --> 0:04:31.360 right And I'll give you an example. 0:04:31.120 --> 0:04:33.679 That I mean, when they're comparing scans, right, it's something 0:04:33.720 --> 0:04:37.599 that might be overlooked. Again, not dissing anybody professionally, but 0:04:38.279 --> 0:04:41.720 when you've got just hardcold data, that's what they're looking at. 0:04:42.200 --> 0:04:44.919 In the study, it'll say roughly half of the physicians 0:04:44.920 --> 0:04:47.160 that use AI as a as a buddy or as 0:04:47.200 --> 0:04:49.839 a check or actually more confident in the ability to 0:04:49.839 --> 0:04:52.080 make their own diagnosis because of the buddy check and 0:04:52.160 --> 0:04:54.360 a quarter of them are saying that it's a significant 0:04:54.560 --> 0:04:57.560 improvement in there in the result that they deliver. They're 0:04:57.600 --> 0:05:01.760 missing they're not missing medical issues, they're not misidentifying, and 0:05:01.800 --> 0:05:04.480 they're not missing things. And that's from the physicians themselves 0:05:04.480 --> 0:05:06.800 that are saying it's actually helping them have higher quality 0:05:06.839 --> 0:05:10.240 diagnostic which if you're dealing with something like cancer potential, 0:05:10.600 --> 0:05:11.440 this is a big deal. 0:05:11.960 --> 0:05:15.120 How do you kind of, I guess, help the physicians 0:05:15.240 --> 0:05:18.040 who maybe see, you know, an introduction of a new 0:05:18.040 --> 0:05:20.920 technology as just another thing that they have to deal 0:05:21.000 --> 0:05:24.680 with on top of their already busy day, Like do 0:05:24.720 --> 0:05:27.120 you get pushed back from maybe I'm trying to think 0:05:27.160 --> 0:05:29.640 of some of those old school doctors who still you know, 0:05:29.680 --> 0:05:32.200 do everything by hand. They still have paper notes, they 0:05:32.200 --> 0:05:34.880 still have all their files, you know, in a pay 0:05:34.920 --> 0:05:37.120 per form, this is a pretty advanced technology if we 0:05:37.200 --> 0:05:37.719 have to implement. 0:05:38.240 --> 0:05:40.000 I go back to this triple threat, the fact that 0:05:40.000 --> 0:05:41.680 there are not enough people to do the work today 0:05:41.960 --> 0:05:43.400 and they're burning out the ones that are doing it, 0:05:43.440 --> 0:05:45.279 And so we actually find in most of the larger 0:05:45.279 --> 0:05:48.479 health systems there's a huge appetite to embrace this. If 0:05:48.760 --> 0:05:52.080 we're doing the design along the clinical workflow. The challenge 0:05:52.160 --> 0:05:54.720 is we took in the industry typically have thrown technology 0:05:54.720 --> 0:05:57.240 of people and said digest it and the workflow has 0:05:57.240 --> 0:05:59.360 to change. We're not doing that at Phillips. We walk 0:05:59.360 --> 0:06:02.800 alongside our our customers. I work with most of all 0:06:02.839 --> 0:06:05.599 the largest health systems in the US and Canada and 0:06:05.640 --> 0:06:07.840 we're walking alongside them. So when we're designing it, we're 0:06:07.839 --> 0:06:11.039 having them in mind. The beauty of AI driven tooling 0:06:11.320 --> 0:06:14.080 is that it can actually do the work if in 0:06:14.120 --> 0:06:16.320 the process that they're in, we can adapt to do 0:06:16.320 --> 0:06:18.760 the work. That's a lot different than deploying software and 0:06:18.839 --> 0:06:21.160 complex things that you have to learn, screens and all that. 0:06:21.680 --> 0:06:24.400 All that goes away and that's really the promise that's 0:06:24.440 --> 0:06:27.200 not this magic black box thing that's actually happening today 0:06:27.200 --> 0:06:29.080 and some of the leading else's. 0:06:28.760 --> 0:06:30.159 So where do you see it all going? Like what 0:06:30.240 --> 0:06:33.360 are you in terms of generative AI and how what 0:06:33.480 --> 0:06:35.960 role do you see in like I don't know in 0:06:36.000 --> 0:06:37.640 the next three to five years, Like do you have 0:06:37.680 --> 0:06:41.080 a feel of how dramatic the impact is well? I think, 0:06:41.720 --> 0:06:43.159 or what are you hearing from the doctors that you 0:06:43.200 --> 0:06:43.520 talk to. 0:06:44.279 --> 0:06:47.080 I think it's a journey that will be rightly done. 0:06:47.279 --> 0:06:52.000 With some conservativism. Again, the operational aspects of clinical practice 0:06:52.160 --> 0:06:55.160 will have dramatic improvement. That's going to give staff back 0:06:55.240 --> 0:06:57.320 time to be with patients, and you're going to get 0:06:57.320 --> 0:06:59.560 a better workforce, and you may even draw people into 0:06:59.800 --> 0:07:02.520 the practice because they actually have these tools that help 0:07:02.560 --> 0:07:05.240 them be better at what they do and they continue 0:07:05.279 --> 0:07:09.320 to have better balance. The sky's the limit, But we 0:07:09.720 --> 0:07:11.480 want to make sure when it comes to generative AI, 0:07:11.560 --> 0:07:15.360 that were really thoughtful about not getting over our skis 0:07:15.400 --> 0:07:18.240 and allowing hallucinations because trust is the number one thing. 0:07:18.320 --> 0:07:20.080 If we build trust with clinicians and we do it 0:07:20.080 --> 0:07:22.880 in the right way, that embrace and that self check 0:07:23.000 --> 0:07:25.800 process actually improves the quality of the output. I've met 0:07:25.800 --> 0:07:28.240 with a bunch of CEOs here in New York last night. Yeah, 0:07:28.320 --> 0:07:30.960 really really top notch institutions, and they're actually telling us 0:07:30.960 --> 0:07:35.280 in six months the quality of the generative AI algorithms 0:07:35.280 --> 0:07:38.160 that they're building is improving itself at a rate they 0:07:38.160 --> 0:07:41.560 didn't anticipate. So again, I think, do it where there's 0:07:41.640 --> 0:07:44.640 high trust, build that trust with clinicians, design it around 0:07:44.640 --> 0:07:47.160 the workflow, and then you'll really get better adoption. And 0:07:47.160 --> 0:07:49.440 we're seeing that we're seeing year to year, we're seeing 0:07:49.440 --> 0:07:52.559 a massive improvement and adoption of tooling that is giving 0:07:52.640 --> 0:07:55.280 life back to a community that joined the practice to 0:07:55.320 --> 0:07:58.400 serve other people. So it's exciting time. 0:07:58.560 --> 0:08:01.520 Okay, So in maybe three to five years, is the 0:08:02.080 --> 0:08:05.360 artificial intelligence actually going to be making clinical decisions? 0:08:05.480 --> 0:08:05.680 I e. 0:08:05.840 --> 0:08:09.480 You can go as a patient into a doctor's office 0:08:09.520 --> 0:08:12.560 and instead of actually seeing the doctor, there's there's not 0:08:12.720 --> 0:08:15.640 a human involved. Essentially the AI is giving you the 0:08:15.680 --> 0:08:16.800 diagnosis and telling you. 0:08:17.200 --> 0:08:19.680 I think it would be foolhardy to think about what 0:08:19.760 --> 0:08:21.680 could be in three to five years because the rate 0:08:21.720 --> 0:08:23.560 of innovation is so fast. What I can tell you 0:08:23.640 --> 0:08:27.320 is we believe as a technology company that decisions always 0:08:27.440 --> 0:08:29.640 land with the physicians, and so we want to augment 0:08:29.720 --> 0:08:31.800 that physician with as much of the non value add 0:08:31.840 --> 0:08:34.120 work and as much of the diagnostic quality they can 0:08:34.200 --> 0:08:36.640 so that they can make the best decision clinically for 0:08:36.679 --> 0:08:39.480 their patients. And I wouldn't I can't see how that'll 0:08:39.559 --> 0:08:41.320 change in the near future. 0:08:41.520 --> 0:08:43.480 Jeff, What does it mean for the devices that you guys, 0:08:43.520 --> 0:08:46.959 whether it's MRIs or cat scans or sonog like the 0:08:46.960 --> 0:08:49.960 device is the medical equipment? How does it impact you 0:08:50.040 --> 0:08:53.160 guys as a company, and how you have to think about, Yeah, 0:08:53.200 --> 0:08:56.280 well we're borating AI or is it kind of pretty easy? 0:08:56.600 --> 0:08:59.080 It's interesting we've made the step to become more of 0:08:59.080 --> 0:09:01.160 a productivity company than a product company. I mean, that's 0:09:01.160 --> 0:09:03.559 a pretty big leap in this business. We think the 0:09:03.600 --> 0:09:08.400 beauty of orchestrating a workflow for healthcare concerns is where 0:09:08.400 --> 0:09:10.440 we want to be. It's not about the products anymore. 0:09:10.440 --> 0:09:12.439 And so when we think of like our patient monitoring journey, 0:09:12.440 --> 0:09:15.600 we're thinking about how do I have a monitoring, persistent 0:09:15.679 --> 0:09:18.920 monitoring experience across every aspect of where a patient could 0:09:18.920 --> 0:09:21.560 be in a healthcare system, even at home. And so 0:09:21.640 --> 0:09:26.640 we're building these platforms around radiology, around cardiology and cardiovascular intervention, 0:09:26.720 --> 0:09:29.600 and around patient monitoring from home to hospital to home, 0:09:29.920 --> 0:09:33.479 where we give hospitals that rich data and that physiological 0:09:33.559 --> 0:09:36.000 data from patients. And we think of it as the 0:09:36.000 --> 0:09:38.440 software layer or the platform that you can start to 0:09:38.440 --> 0:09:43.000 build these AI algorithms into exactly right, think about the 0:09:43.000 --> 0:09:45.760 apps on your iPhone. Oversimplified, but think about apps on 0:09:45.800 --> 0:09:46.240 your iPhone. 0:09:46.280 --> 0:09:46.400 Right. 0:09:46.440 --> 0:09:48.760 You plug the apps and that do certain things, but 0:09:48.840 --> 0:09:51.880 the platform itself and that's where we've really moved our 0:09:51.960 --> 0:09:52.199 R and D. 0:09:52.440 --> 0:09:54.120 Hey listen, we'd be a remissed you've only got about 0:09:54.120 --> 0:09:55.439 a minute and a half left here, but I got 0:09:55.480 --> 0:09:58.080 to ask you about the macro You see a lot. 0:09:58.360 --> 0:10:01.360 Your company obviously is global charge of the North American unit. 0:10:02.520 --> 0:10:05.160 How would you describe the business environment and the consumer 0:10:05.240 --> 0:10:05.839 side of things? 0:10:06.840 --> 0:10:08.800 Actually pretty strong for both, at least for Phillips. In 0:10:08.840 --> 0:10:11.760 the consumer segments we're in, we're seeing double digit growth 0:10:11.760 --> 0:10:15.200 across the globe, particularly strong here in North America. Consumer 0:10:16.120 --> 0:10:17.920 demand has actually gone up for us, and we've got 0:10:17.920 --> 0:10:22.240 a great portfolio. In the health system side, again, we've 0:10:22.240 --> 0:10:26.640 seen double digit growth last year, where we've consistently outpaced 0:10:26.640 --> 0:10:30.200 the industry for some time. But about the industry itself, 0:10:30.280 --> 0:10:32.040 we see strong demand because of all the things we've 0:10:32.080 --> 0:10:34.520 been talking about. We're pivoting to be less about the 0:10:34.520 --> 0:10:36.400 product and more about productivity. And you think that's going 0:10:36.440 --> 0:10:38.160 to be a game changer for health systems. That's long 0:10:38.240 --> 0:10:38.760 term demand. 0:10:39.120 --> 0:10:40.959 How are you guys using AI at your company? 0:10:41.320 --> 0:10:43.960 So we're deploying everything from our financial systems to my 0:10:44.000 --> 0:10:47.480 commercial operations, to contract management, our services portfolio so that 0:10:47.520 --> 0:10:51.440 we can deploy agents to look at systems health systems 0:10:51.760 --> 0:10:53.520 products that we have in the field to be able 0:10:53.520 --> 0:10:58.600 to do automatic determination of root cause and a correction path. 0:10:58.920 --> 0:11:03.800 We're as committed internally to driving productivity and efficiency so 0:11:03.840 --> 0:11:07.559 that we can grow with our customers at at pace. 0:11:07.679 --> 0:11:10.720 Yeah, it's kind of falling. I know, I know, I 0:11:10.760 --> 0:11:11.320 know your. 0:11:11.240 --> 0:11:11.800 Own dog food. 0:11:11.800 --> 0:11:15.480 That's right, that's exactly Jeff Delulo, he is CEO at 0:11:15.480 --> 0:11:18.320 Phillips North America. Thank you so much. Really appreciate it. Yeah, 0:11:18.360 --> 0:11:19.040 we appreciate it.