Latest / Best of LinkedIn: Digital Transformation & Tech / Best of LinkedIn: Health Tech CW 40/ 41

Transcript

Best of LinkedIn: Health Tech CW 40/ 41 · Best of LinkedIn: Digital Transformation & Tech

• 0:00
  Speaker 0: This episode is provided by Thomas Allgaier and Frenius, based on the most relevant LinkedIn posts about health tech from CW-Forty in Forty One.
• 0:07
  Speaker 0: Frenias equips HealthTech providers with The Market Intelligence to identify which hospitals to target... ...and how to reach decision makers for hospital digitalization as a result of the Krankenhauszukumskisatz.
• 0:19
  Speaker 0: You can find more info in the description.
• 0:21
  Speaker 1: So welcome to the deep dive.
• 0:24
  Speaker 0: Yeah thanks for joining us Today.
• 0:26
  Speaker 0: we are digging into the really big health tech trends.
• 0:29
  Speaker 0: We've seen floating across LinkedIn recently
• 0:31
  Speaker 1: and we are skipping The hype here, we really want to focus on How data in architecture an actual business models?
• 0:38
  Speaker 1: Are shifting on the ground
• 0:39
  Speaker 0: exactly.
• 0:40
  Speaker 0: I mean just at this stage.
• 0:42
  Speaker 0: Imagine you're checking your new lab results on a nap right And your personal AI doctor pops up and says hey start taking A daily aspirin
• 0:50
  Speaker 1: which sounds totally normal.
• 0:51
  Speaker 1: right
• 0:51
  Speaker 0: it's sound standard except Your real human doctor actually took you off aspirin years ago because based on your specific medical history, taking it could literally cause severe internal bleeding.
• 1:05
  Speaker 0: Oh
• 1:05
  Speaker 1: yeah!
• 1:05
  Speaker 1: That is terrifying.
• 1:07
  Speaker 0: It IS and that happened recently to a healthcare leader.
• 1:12
  Speaker 0: So today we're cutting through the marketing noise to look at why Healthcare AI is well kind of broken right now And how tech giants are physically re-architecting systems.
• 1:22
  Speaker 1: Yeah, that aspirin story really captures the exact moment we're in right now.
• 1:26
  Speaker 1: I mean... We are officially past the whole what-if phase of healthcare tech.
• 1:29
  Speaker 0: Definitely
• 1:30
  Speaker 1: Because everyone has a model Now Right But very few actually have a genuinely trustworthy system.
• 1:35
  Speaker 1: Tarek Matty lead this out beautifully on LinkedIn.
• 1:38
  Speaker 1: He argued that trustworthy health care AI isn't just you know A vague value statement.
• 1:43
  Speaker 1: You can slap on pitch deck.
• 1:44
  Speaker 0: It actually means something?
• 1:46
  Speaker 0: Right!
• 1:46
  Speaker 1: It demands four Really concrete checkable things A named owner abounded claim, human oversight and a measured benefit against the real baseline.
• 1:58
  Speaker 0: Wait I'm noticing pretty glaring omission from that list like where's accuracy?
• 2:02
  Speaker 0: because usually thats very first metric vendors throw at you when they're pitching tool?
• 2:07
  Speaker 1: I know but that omission is actually entirely intentional Because Accuracy really only describes how well your algorithm performed on a sanitized historical test set
• 2:17
  Speaker 0: Right Like in lab setting
• 2:18
  Speaker 1: Exactly.
• 2:19
  Speaker 1: It tells literally nothing about whether the system can survive the chaos of a real clinical floor where data is messy, patients are unpredictable and things change every single day.
• 2:29
  Speaker 1: Accuracy is just a snapshot.
• 2:32
  Speaker 1: Trustworthiness is a continuous operational
• 2:35
  Speaker 1: state.".
• 2:35
  Speaker 0: So let's push on that operational state for second especially as tech is evolving so fast because having human double check or static recommendation it one thing.
• 2:45
  Speaker 0: but what happens when AI autonomous?
• 2:49
  Speaker 0: Like, Marcus Schmidt brought up agentic AI and it feels like the threat surface completely changes there.
• 2:55
  Speaker 0: When an AI agent can read a record call in external tool And just act on result without asking for permission The potential of catastrophic failure?
• 3:04
  Speaker 0: Skyrockets.
• 3:05
  Speaker 1: Oh wait absolutely skyrockets If you are still treating it as standard software rollout.
• 3:11
  Speaker 1: Schmitt argues that this shift basically requires compliance & security.
• 3:16
  Speaker 1: by design
• 3:17
  Speaker 0: Right
• 3:18
  Speaker 1: We're talking about scoped identities for individual AI agents, permissions that explicitly expire after a task is done and engineered.
• 3:26
  Speaker 1: hard stops where the system just physically cannot proceed without a human clearing it.
• 3:31
  Speaker 0: Well let's break down the mechanics of those hard stops because usually protecting protected health information or PHI relies on written policies right?
• 3:41
  Speaker 0: we tell developers hey here are the rules.
• 3:43
  Speaker 0: please don't send this patient data to an external server.
• 3:46
  Speaker 0: But a policy is really just an agreement and agreements drift.
• 3:51
  Speaker 1: And that's exactly why Victor Rodriguez pointed out, you know with over sixteen hundred FDA authorized AI devices currently out in the wild The primary risk isn't an AI model hallucinating In an isolated bubble.
• 4:03
  Speaker 0: No it's connections
• 4:04
  Speaker 1: Exactly!
• 4:05
  Speaker 1: The risk is in the connection between all those systems where data transforms and ownership gets blurry.
• 4:12
  Speaker 1: Healthcare AI is fundamentally an architecture problem now, not just a math
• 4:16
  Speaker 0: problem.
• 4:17
  Speaker 0: Right and Deepak Upade illustrated how we actually solve this using Oracle Cloud Infrastructure OCI.
• 4:24
  Speaker 0: you don't Just put up You know do-not-steal sign for the AI.
• 4:28
  Speaker 0: use native controls to run The models in a subnet where?
• 4:31
  Speaker 0: The route table literally has zero rules For outbound traffic.
• 4:36
  Speaker 1: right.
• 4:36
  Speaker 1: so it's Not just a policy saying Don't take the data
• 4:38
  Speaker 0: exactly.
• 4:39
  Speaker 0: It's like building a room with no doors.
• 4:41
  Speaker 0: You physically cannot leave because there is no exit route built into the architecture.
• 4:45
  Speaker 0: The data, physically can not leak.
• 4:47
  Speaker 1: Yeah and that structural limitation Is really the only way to guarantee security at scale.
• 4:52
  Speaker 1: Alson Brunel noted That four classic questions of Data Governance They haven't changed in thirty years.
• 4:58
  Speaker 1: But agentic AI makes them so incredibly urgent.
• 5:02
  Speaker 0: Like who owns the data?
• 5:04
  Speaker 0: Where it is Who has access And is accurate
• 5:07
  Speaker 1: Exactly?
• 5:07
  Speaker 1: And we always have to circle back the ultimate boundary line here.
• 5:11
  Speaker 1: Jared L. Skillings and Charlene Lee, she's actually one who experienced that rogue aspirin recommendation We talked about earlier.
• 5:16
  Speaker 1: Oh wow!
• 5:17
  Speaker 0: That was her.
• 5:18
  Speaker 1: Yeah They both emphasize while you can delegate tasks To an AI You could never ever Delegate clinical responsibility.
• 5:26
  Speaker 0: Well, the terrifying part of that story is what would have happened if she hadn't known her own medical history well enough to push back?
• 5:32
  Speaker 0: I know.
• 5:32
  Speaker 0: Because when she actually challenged the AI it reevaluated her chart and then agreed that the bleeding risk outweighed the benefit.
• 5:39
  Speaker 0: but proving an A.I.. Is safe inside a room with no doors as just step one
• 5:44
  Speaker 1: right totally.
• 5:46
  Speaker 0: How do you actually convince a hospital executive to buy the thing?
• 5:49
  Speaker 0: because if architecture secures The AI how do you vendors prove It Actually works To the buyers?
• 5:55
  Speaker 1: The bar for that has become just ruthless.
• 5:58
  Speaker 1: buyers are demanding concrete clinical and financial evidence before they even entertain a conversation right now.
• 6:04
  Speaker 1: And frankly, a lot of vendors.
• 6:05
  Speaker 1: They're just failing the test.
• 6:07
  Speaker 0: I mean take predictive AI For example.
• 6:09
  Speaker 0: Elena Turcott offered a really sharp critique Of what's currently on the market?
• 6:13
  Speaker 0: She says most predictive medical AI isn't actually predicting anything at all.
• 6:17
  Speaker 1: Right
• 6:18
  Speaker 0: It's just very expensive early detection.
• 6:21
  Speaker 1: Yeah, the mechanism here is so vital to understand because most of these models are trained to spot existing signals right?
• 6:27
  Speaker 1: Like subtle voice changes or microscopic inflammatory markers.
• 6:32
  Speaker 1: But spotting a biological process that has already started Is not forecasting.
• 6:38
  Speaker 1: True forecasting requires an actual causal structure.
• 6:42
  Speaker 1: You have to mathematically model how biological and systemic variables interact over time rather than just pointing at a correlation that's already materialized.
• 6:51
  Speaker 0: And even when the tool does exactly what it promises, like... What is actually changing?
• 6:57
  Speaker 0: Bargav Patel reviewed forty-eight different studies on large language models used for patient education.
• 7:02
  Speaker 1: Oh!
• 7:02
  Speaker 1: That was fascinating review.
• 7:03
  Speaker 0: Yeah and data shows LLMs definitely improve patients comprehension of their disease but actual behavior change.
• 7:11
  Speaker 0: That remains completely unproven.
• 7:13
  Speaker 0: Like understanding the mechanics of your hypertension doesn't matter to the hospital's bottom line if you still don't take your
• 7:18
  Speaker 0: medication.".
• 7:19
  Speaker 1: And that right.
• 7:20
  Speaker 1: there is the measurement trap in medical AI, right now.
• 7:24
  Speaker 1: Vendors measure what's easy like patient comprehension or time saved.
• 7:28
  Speaker 1: drafting a clinical note and they just assume it naturally translates into what's hard which is better clinical outcomes than lower readmission rates.
• 7:36
  Speaker 0: So put yourself in the shoes of a hospital CEO for a minute.
• 7:40
  Speaker 0: If you are running a health system on a razor thin one to two percent margin, how do you justify buying tech with unproven outcomes?
• 7:47
  Speaker 1: You absolutely don't
• 7:48
  Speaker 0: exactly.
• 7:49
  Speaker 0: Deepika Bodapati observed that these low-margin health systems are flat out refusing to run pilots anymore.
• 7:55
  Speaker 0: They're essentially treating AI vendors the way I treat a new streaming service, letting my friends pay for it and tell me if its any good before i
• 8:01
  Speaker 0: commit.".
• 8:03
  Speaker 0: They wait for rigorous buyers like Tenet Healthcare to run those intense vendor bake offs down at the decimal point once they prove their AI works then smaller systems will actually buy
• 8:16
  Speaker 1: them.
• 8:18
  Speaker 1: Tim Brown notes that vendors get stuck in pilot purgatory because they have unclear buyers, missing budgets or just no real expansion path planned.
• 8:26
  Speaker 0: Right yeah Because selling the first trial use case is totally different from scaling it across an entire network.
• 8:32
  Speaker 1: right
• 8:32
  Speaker 0: and Thomas Gilbertson adds to that saying if you want to create measurable value The pre-development decisions like actually understanding the hospital specific workflow constraints That matters far more than a sophistication of your AI model.
• 8:45
  Speaker 0: but
• 8:45
  Speaker 1: We should balance the scales here a bit because AI is proving massive, measurable value in specific areas when it's deployed correctly.
• 8:53
  Speaker 0: Oh absolutely!
• 8:54
  Speaker 1: Like Mark Shapiro shared method using LLMs that extracted oncology data from electronic health records at over ninety-five percent accuracy and had actually found two hundred percent more distinct therapies than structured data feeds alone.
• 9:07
  Speaker 0: Wow thats huge.
• 9:09
  Speaker 1: And going back to true forecasting, Yossi Matias noted that Google's ERA AI model ranked first among thirty-nine different models in the CDC flu site forecasting for hospital admissions.
• 9:20
  Speaker 0: Okay those are incredibly powerful capabilities but they do carry immense responsibility.
• 9:25
  Speaker 0: Like Parisa Ashar published a framework warning that we really need disability-informed AI.
• 9:32
  Speaker 1: Yes,
• 9:33
  Speaker 0: because if we build triage algorithms using historical data that was shaped by inaccessible care or structural barriers the AIs just going to misclassify patients.
• 9:41
  Speaker 0: it might deny resources to vulnerable populations based on historical bias rather than true clinical needs
• 9:48
  Speaker 1: exactly.
• 9:49
  Speaker 0: By the way, if you are finding this breakdown of the health tech landscape useful so far make sure you subscribe.
• 9:54
  Speaker 0: So You don't miss our future deep dives.
• 9:55
  Speaker 0: We are constantly unpacking that technologies That Are quietly running The world.
• 9:59
  Speaker 1: Speaking Of Running The World Proving That An AI Improves Patient Outcomes Is Completely Useless If A Hospital Isn'T Actually Allowed To Deploy It.
• 10:07
  Speaker 1: Which Brings Us to The Massive Bottleneck That Regulators And Payers Are Currently Fighting Over.
• 10:11
  Speaker 0: Oh, regulation and prior authorization.
• 10:14
  Speaker 0: It is the absolute choke point of health care innovation right now!
• 10:18
  Speaker 1: It really...really IS.
• 10:20
  Speaker 1: Manjushri SP highlighted a major mindset shift that vendors need to make here.
• 10:24
  Speaker 1: Medical Device Regulatory Affairs can no longer be treated as one-time submission task where you just get your stamp and walk
• 10:30
  Speaker 0: away Right.
• 10:31
  Speaker 0: it's not finish line.
• 10:32
  Speaker 1: Exactly it has to become a life cycle function.
• 10:36
  Speaker 1: When you update a machine learning algorithm, You have to constantly measure what that change means for safety?
• 10:41
  Speaker 1: For clinical evidence and cybersecurity throughout the device's entire life.
• 10:46
  Speaker 0: Well, The regulators themselves are actually adapting To this new reality too.
• 10:50
  Speaker 0: David Lobaho Isghierdo pointed out A really fascinating shift with the FDA's tempo program.
• 10:54
  Speaker 0: Oh right Instead of forcing vendors to wait until validation is entirely complete in a vacuum before they can deploy, the FDA is letting selected generative AI tools actually generate clinical evidence through real world deployment.
• 11:08
  Speaker 1: And we're seeing a push for efficiency globally too.
• 11:10
  Speaker 1: Robert Z Phillips noted from The International Medical Device Regulators Forum that regulatory reliance is gaining serious momentum.
• 11:18
  Speaker 1: meaning
• 11:18
  Speaker 0: what exactly?
• 11:19
  Speaker 1: Meaning countries are sharing frameworks.
• 11:23
  Speaker 1: We're also seeing vital standardizations emerge, like Connor Walsh mentioned the health ISAC baselines to secure medical devices against cyber attacks.
• 11:31
  Speaker 1: And Georg Heidenreich detailed expert reviews for IEC six-oh-six zero one safety standards which is basically a foundational rule book for medical electrical equipment.
• 11:41
  Speaker 0: Okay but let's shift from regulators and payers specifically the absolute nightmare of prior authorization.
• 11:50
  Speaker 0: Harry Friedman pointed out that the new CMS, uh, zero-zero five seven APIs.
• 11:55
  Speaker 0: Um these are the rules meant to standardize how data is exchanged for prior off.
• 11:59
  Speaker 0: they are not magic ones.
• 12:00
  Speaker 0: no
• 12:01
  Speaker 1: definitely
• 12:01
  Speaker 0: not.
• 12:02
  Speaker 0: They only reduce the administrative burden if payers actually digitized their clinical policies and If providers structure their clinical data so it can autofill those questionnaires.
• 12:11
  Speaker 1: Right because if you just shift a forty page questionnaire from a clunky payer portal To a clunky EHR screen you haven't actually solved the workflow problem.
• 12:21
  Speaker 0: Exactly,
• 12:23
  Speaker 1: which is why Anusia Trivedi suggests replacing adversarial prior authorization entirely with headless agents.
• 12:30
  Speaker 0: Okay hold on let me play devil's advocate here for a second.
• 12:33
  Speaker 0: Headless agents means there is no human interface right?
• 12:37
  Speaker 0: You want to let two AI systems securely ping data back and forth To decide if my surgery gets approved without a human doctor looking at the screen as A patient.
• 12:47
  Speaker 0: that sounds like A massive security and ethical nightmare.
• 12:51
  Speaker 0: How do we trust
• 12:51
  Speaker 1: that?
• 12:52
  Speaker 1: The apprehension is completely valid, I get it.
• 12:55
  Speaker 0: but
• 12:56
  Speaker 1: look at the alternative We currently endure if we don't use headless APIs.
• 13:00
  Speaker 1: We are essentially using one AI at the insurance company to draft an angry denial letter based on a keyword mismatch
• 13:06
  Speaker 0: right And
• 13:07
  Speaker 1: then the hospital uses another AI to draft a ten-page angry appeal letter.
• 13:11
  Speaker 0: It's basically like arming two lawyers with faster typewriters.
• 13:14
  Speaker 0: Yes The fight is the exact same, it just happening at lightning speed and patient still sitting there waiting for care.
• 13:20
  Speaker 1: Exactly a problem.
• 13:22
  Speaker 1: Headless agents don't argue each other.
• 13:24
  Speaker 1: They securely share actual clinical evidence via APIs to reach an instant binary decision based purely on whether chart data matches digital policy requirements.
• 13:36
  Speaker 1: You remove human from data entry friction not from clinical oversight.
• 13:41
  Speaker 0: Okay, that makes a lot of sense when you frame it like that.
• 13:44
  Speaker 0: And workflow integration isn't just happening in the billing office either.
• 13:47
  Speaker 0: It's physically reshaping and operating room.
• 13:49
  Speaker 0: This is where hardware gets really interesting.
• 13:52
  Speaker 1: Oh definitely
• 13:53
  Speaker 0: Mitra Sultani had great insight regarding Medtronics FDA clearance for their Hugo robotic assisted surgery system.
• 14:02
  Speaker 0: She said something kind changes how view whole industry.
• 14:05
  Speaker 0: The robot not product...the workflow
• 14:10
  Speaker 1: The strategy there is just brilliant, because vessel sealing is a critical high stakes part of surgery.
• 14:16
  Speaker 1: Ligashore is a trusted twenty year old technology that surgeons know intimately
• 14:20
  Speaker 0: Right.
• 14:21
  Speaker 1: So by bringing the deeply familiar tool into brand new robotic arm you completely remove friction and adoption.
• 14:28
  Speaker 1: Surgeons don't have to learn a completely new way to seal vessels while they're also trying master a robot.
• 14:34
  Speaker 0: And business model behind this ecosystem approach is staggering.
• 14:38
  Speaker 0: Michelle Bohan broke down the numbers for Intuitive, The makers of the famous Da Vinci surgical robot.
• 14:43
  Speaker 0: Eighty-four percent of their revenue comes from instruments, accessories and services... Not from selling giant robots themselves!
• 14:51
  Speaker 0: In fact they recently placed nearly five hundred systems on usage based leases.
• 14:56
  Speaker 0: You don't buy a robot you pay per procedure.
• 14:58
  Speaker 0: A machine is just the meter…the money in utilization.
• 15:03
  Speaker 1: But as capital floods into this space and the volume of procedures rises.
• 15:08
  Speaker 1: The push for actual surgical autonomy, like letting the robot make cuts is accelerating
• 15:13
  Speaker 1: fast."
• 15:14
  Speaker 0: Which brings up a massive safety question!
• 15:17
  Speaker 0: With all this money flying around how do we ensure patient's safety isn't lost in pursuit to having a robot?
• 15:21
  Speaker 0: do all that work?
• 15:23
  Speaker 0: Amadou Weyordeau offered the perfect metric for it – he said surgical autonomy shouldn't be judged by what the robot can do but by how safely human can take back control because surgery is unpredictable.
• 15:35
  Speaker 0: Knowing exactly when the system needs to stop and hand control back to the human surgeon is infinitely more important than automating some complex cut.
• 15:43
  Speaker 1: And ensuring that safe handoff requires intense, continuous training... ...and super precise pre-surgical planning.
• 15:51
  Speaker 1: The experts across the industry really agree on this.
• 15:53
  Speaker 1: We're seeing a push for proficiency based progression To make robotic surgery training measurable Like Bob Reddy discussed And Jim Patient and Rajesh Kumar talked about platforms like Metronics Able, which builds patient-specific spine plans from MRI images.
• 16:10
  Speaker 1: Before the patient is even in the room...
• 16:12
  Speaker 0: It's amazing!
• 16:13
  Speaker 1: ...the precision is reaching incredible levels.
• 16:17
  Speaker 1: We're now seeing radiotherapy solutions as Burnt Montag pointed out with Xcela that can treat multiple tumors within a single patient breath hold.
• 16:25
  Speaker 0: A single breath hold?
• 16:27
  Speaker 1: Yeah
• 16:27
  Speaker 0: The scale of that precision.
• 16:28
  Speaker 0: it just hard to wrap your head around But precision technology isn't staying trapped inside the hospital.
• 16:34
  Speaker 0: It is scaling out into outpatient diagnostics and directly in to consumers hands.
• 16:39
  Speaker 1: Yeah, it's definitely happening at advanced imaging first.
• 16:42
  Speaker 1: Catarina Hessels shared a really striking clinical case involving photon counting CT.
• 16:47
  Speaker 0: Okay
• 16:48
  Speaker 1: This technology revealed as C-II dense fracture which was tiny fracture on upper spine that conventional CT scan completely missed.
• 16:56
  Speaker 0: Wow
• 16:57
  Speaker 1: The clinical impact is so massive that burned ownership noted Siemens health and ears had to build a brand new sensor fab in Forheim just to handle the global demand for this next generation imaging.
• 17:07
  Speaker 0: So let's unpack the mechanism there.
• 17:10
  Speaker 0: How does photon counting actually differ from the CT scanners hospitals have been using for decades?
• 17:15
  Speaker 0: like what?
• 17:16
  Speaker 0: Is it physically doing differently to see a tiny fracture that a normal scan misses?
• 17:21
  Speaker 1: well traditional CT scammers use detectors that basically measure the total energy of all the X-rays hitting them at once.
• 17:29
  Speaker 1: It's kind of like trying to measure rainfall by only looking at a puddle, you just get a blurry aggregate.
• 17:34
  Speaker 1: but photon counting CT as the name applies it literally counts every single individual x-ray photon hitting the sensor.
• 17:42
  Speaker 1: It measures the exact energy level of each particle.
• 17:45
  Speaker 0: so instead of a blurry puddle you are measuring every single drop of rain?
• 17:50
  Speaker 0: The noise is eliminated
• 17:51
  Speaker 1: exactly!
• 17:53
  Speaker 0: But the scale gets even wilder when we look.
• 17:56
  Speaker 0: Keaton Badell pointed out a massive shift in how virtual care is delivered.
• 18:00
  Speaker 0: Grinder just made a two hundred and fifty million dollar deal to acquire Freddie, which as the telehealth company focused on PrEP.
• 18:07
  Speaker 1: I mean this completely rewrites economics of Virtual Care.
• 18:10
  Speaker 0: It really does Because consumer platforms are solving Virtual Care's hardest most expensive problem Which is patient acquisition.
• 18:17
  Speaker 0: Yep You can build the most secure advanced TeleHealth Clinic In The World But finding patients And getting them To download your app Is incredibly Expensive.
• 18:25
  Speaker 0: Brands with massive highly engaged audiences are becoming the new health care clinics because they already have The audience checking their phones ten times a day.
• 18:33
  Speaker 1: Don Q. Dow is seeing this exact thing play out with NOLA Health.
• 18:37
  Speaker 1: They're piloting an end-to-end AI dermatology pathway in Utah for adult acne.
• 18:42
  Speaker 1: It handles the entire life cycle from intake to image analysis To the initial prescription all the way through to refills.
• 18:49
  Speaker 0: that's incredible
• 18:50
  Speaker 1: and tech giants are leaning in heavily too.
• 18:52
  Speaker 1: Andres Ponce noted the revamped Apple Health app will soon calculate a health age, and readiness score synthesizing your sleep activity data into actionable clinical insights.
• 19:03
  Speaker 0: Yeah, Dr.
• 19:04
  Speaker 0: Ben Marathapu and Chris Chesney both celebrated time naming CERA and Teladoc Health as top health tech companies of the year.
• 19:11
  Speaker 0: I just proves that deploying AI to the front line of home care and virtual care is scaling rapidly.
• 19:16
  Speaker 0: It's creating more space for human care by automating the administrative friction out-of-the process
• 19:21
  Speaker 1: Totally!
• 19:22
  Speaker 1: If we synthesize all of the developments from these last two weeks The core takeaway is pretty undeniable The winner in healthtech isn't a startup with best standalone algorithm.
• 19:32
  Speaker 1: The winners are the ecosystems, whether it's Oracle securing the data architecture in a room with no doors or headless agents silently solving the prior off workflow.
• 19:44
  Speaker 1: Or consumer apps owning the daily patient relationship.
• 19:47
  Speaker 1: success is now entirely about how well you connect to pieces.
• 19:51
  Speaker 0: so as we wrap up I want to leave You With A Final Thought To Chew On As These Consumer Platforms With Massive Highly Engaged Audiences Become The New Front Door To Virtual Care.
• 20:01
  Speaker 0: How will traditional health systems compete for patient loyalty when a tech brand already owns their daily habits?