Latest / HumanifAI HealthTech / How AI is Reimagining the EHR w/ Shweta Maniar, Global Director, Google Cloud
Transcript
- 0:00SPEAKER_00: The EHR was supposed to fix healthcare.
- 0:03SPEAKER_00: Instead, it created a new crisis, one that burns out the very clinician that it was meant to help.
- 0:09SPEAKER_00: Now, today, a leader from Google Cloud is here to show us what it actually looks like when AI finally delivers on that original promise.
- 0:35SPEAKER_00: I'm your host, Francesca Vereb, Head of Marketing at NetHealth, and today we are thrilled to welcome Shweta Maynard, Global Director of Healthcare and Life Sciences Solutions and Strategy at Google Cloud.
- 0:49SPEAKER_00: Now, Shweta has spent her career turning complex technology into something healthcare organizations can actually use.
- 0:56SPEAKER_00: And at Google Cloud, she works with some of the world's most sophisticated healthcare systems, moving the AI concept to clinical reality.
- 1:05SPEAKER_00: Before Google, she was actually a Gen Entech, where she built healthcare technology accelerators from the ground up and evaluated thousands of digital healthcare startups for acquisition and investment.
- 1:17SPEAKER_00: And she's a sought-after voice at HIMS, Health, and Google Cloud Next, and was named Fierce Healthcare's Top 10 Woman of Influence in 2024.
- 1:27SPEAKER_00: Congratulations, Shweta.
- 1:29SPEAKER_00: We're especially excited to have her here today ahead of our session at NetHealth's 2026 Next Conference, which is happening in April.
- 1:37SPEAKER_00: And I have a feeling that this conversation is going to set the table perfectly for what's coming.
- 1:42SPEAKER_00: Shweta, welcome to the show.
- 1:44SPEAKER_00: Thank you so much for having me, Francesca.
- 1:47SPEAKER_00: Well, I'm delighted.
- 1:48SPEAKER_00: I have been prepping for this session, and I am so excited.
- 1:50SPEAKER_00: I have a long list of great questions for you.
- 1:53SPEAKER_00: So let's jump right in.
- 1:55SPEAKER_00: Starting with EHRs, right?
- 1:57SPEAKER_00: So the EHR promised to be the single source of truth that would improve outcomes and free clinicians up to focus on patients.
- 2:05SPEAKER_00: But an AMA survey in 2024 found that nearly two-thirds of the physicians actually say EHRs contribute to burnout.
- 2:12SPEAKER_00: Now Schweta, you work with healthcare systems across the world at Google Cloud.
- 2:16SPEAKER_00: And before that, you spend years evaluating thousands of digital companies trying to solve exactly that problem.
- 2:23SPEAKER_00: Where did the original EHR's promise break down and how is AI being used to help improve it?
- 2:30SPEAKER_01: Well, so that's the core of it, isn't it?
- 2:32SPEAKER_01: Right?
- 2:32SPEAKER_01: That's the core of the issue.
- 2:33SPEAKER_01: The promise of these um health records was really to create that single source of information that's going to change the way that we would have care.
- 2:47SPEAKER_01: But for so many of these care providers, it ended up becoming uh another form of burnout, another form of burden, another thing that had to be done.
- 3:00SPEAKER_01: And so the promise actually, I would say, broke down because the first generation of these records ended up becoming uh a digital form of these kind of uh filing cabinets, right?
- 3:15SPEAKER_01: And these billing systems.
- 3:16SPEAKER_01: So they weren't actually built as I would call them clinical workflows in mind, um, because they weren't actually designed to make these massive amounts of data useful, right?
- 3:28SPEAKER_01: And useful at the point of care.
- 3:30SPEAKER_01: And so I would say that we're at a different moment right now, Francesca, because now we have the ability to break down data silos that have actually plagued the healthcare system, the entire healthcare and life science ecosystem, uh for decades.
- 3:46SPEAKER_01: So now we have the ability to break down these silos and actually make this information much more usable.
- 3:51SPEAKER_01: Uh now we have AI, specifically generative AI, which we'll talk about a little bit more, I'm sure.
- 3:57SPEAKER_01: Um, and so now we're moving beyond just machine learning to large language models and now what we call agentic AI.
- 4:04SPEAKER_01: So it's not just that these are incremental improvements.
- 4:08SPEAKER_01: Now we're looking at a paradigm shift completely.
- 4:11SPEAKER_01: And so um at Google, we're seeing this transformation firsthand with partners uh like HCA Healthcare, um, where we're seeing people streamline uh workflows uh to give time back to the healthcare providers.
- 4:26SPEAKER_01: And so finally we're starting to see organizations are really being able to deliver on the original promise of using these kind of tools like records, healthcare records, to deliver on the original power, on the original promise of being able to empower healthcare providers so that they can do what they are, what they should be focusing on, which is help care for patients.
- 4:51SPEAKER_00: Such a fascinating point.
- 4:53SPEAKER_00: And just one quick moment, right?
- 4:55SPEAKER_00: So leaders across the world are being asked to make these major investments into technology, and AI terminology keeps getting shifted around within this consideration, right?
- 5:05SPEAKER_00: Things, terms like machine learning, uh, large language models, you mentioned yourself, agentic AI, but these get conflated all the time.
- 5:14SPEAKER_00: But the distinctions have really big consequences for how organizations plan and spend.
- 5:19SPEAKER_00: You touched a little bit on that yourself in this whole consideration for EHRs.
- 5:22SPEAKER_00: Now, in our conversations, prepping for this session, but also in all the talks that you've done, you draw really clear distinctions between the different types of AI.
- 5:33SPEAKER_00: So, why does knowing the difference really matter from a practical sense for a clinician clinician or for an operational leader who might be deciding where they should put their resources?
- 5:44SPEAKER_01: You know, when we're talking about machine learning, we're talking about a broad field of artificial intelligence where we have algorithms that are being trained on data to identify patterns and predictions on decisions that without being explicitly programmed on a tr on a task.
- 6:06SPEAKER_01: So there's supervised machine learning and unsupervised machine learning.
- 6:09SPEAKER_01: And as it is called, supervised is considered to be trained on data that's specifically made to predict outcomes.
- 6:16SPEAKER_01: And unsupervised machine learning is uh trained on unlabeled data to find hidden patterns.
- 6:22SPEAKER_01: So that's what machine learning is.
- 6:24SPEAKER_01: And large language models or LLMs are what I would consider a specific subset of machine learning or deep learning and deep learning.
- 6:33SPEAKER_01: And LLMs are actually models that are trained on massive amounts of text data.
- 6:38SPEAKER_01: So if you think about large language models in the context of healthcare, these are text data that are meant to understand and generate and interact in a human-like way.
- 6:48SPEAKER_01: So when we're talking about LLMs in a healthcare context, we're talking about being able to understand how to interact with medical information, right?
- 6:58SPEAKER_01: So they're generally considered to be reactive.
- 7:00SPEAKER_01: How do you respond to prompts?
- 7:02SPEAKER_01: So this is in a healthcare context, it's like summarizing clinical notes, patient histories, answering medical questions.
- 7:09SPEAKER_01: And now the last one that you mentioned, right, which is about agentic AI, that actually represents the most advanced implementation of AI.
- 7:19SPEAKER_01: That's actually building on large language models by adding a level of autonomy.
- 7:25SPEAKER_01: So if you take large language models and then you add a layer of autonomy, now an AI agent can actually somewhat independently plan and execute multi-steps to achieve a specific goal.
- 7:39SPEAKER_01: So if you were going to talk about in a healthcare context, an agentic AI and healthcare could potentially automate administrative tasks like scheduling, perhaps uh assisting in prior authorizations, engaging with patients for follow-up scheduling or monitoring, things like that.
- 7:58SPEAKER_01: And so this is why we're that this is why I think it's very important, Francesca, that we all need to understand and take the time to understand the difference between when we're talking about LLMs versus agentic AI, because these are all important advancements in this world that we're living in when it comes to machine learning and deep learning.
- 8:17SPEAKER_00: That's fantastic.
- 8:18SPEAKER_00: So I'm gonna press in just to make sure I understand and then also take it maybe one step further.
- 8:24SPEAKER_00: So pressing in specifically on agentic AI, because I agree with you, that's that is the most exciting, also the most powerful.
- 8:32SPEAKER_00: So that is about generating more interest.
- 8:34SPEAKER_00: Uh sorry, not about generating more interest, it's a it's generating more interest, but it's also creating that confusion, as you were saying.
- 8:41SPEAKER_00: And unlike uh ambient documentation as an example, which is simply listening and transcribing, as you said, the whole point of agentic AI is that it is goal-directed, right?
- 8:53SPEAKER_00: And that it actually then is able to not just capture, but start to propose some next steps and begins acting while still keeping the human in the loop.
- 9:02SPEAKER_00: And one of the things I found fascinating in the research that I read that you had done, um, you were showing that nearly half of healthcare executives already have AI agents running in production.
- 9:13SPEAKER_00: So tell us a little bit more about that specifically.
- 9:16SPEAKER_00: For a physical therapist or a wound care nurse, for example, on it on their daily rounds, what does agentic healthcare AI actually look like in practice?
- 9:25SPEAKER_01: So for too long, AI has felt very abstract, particularly when we're talking about on the front lines.
- 9:31SPEAKER_01: And agentic AI could change that.
- 9:33SPEAKER_01: And I think this is where you can think of it as a proactive partner.
- 9:37SPEAKER_01: Human is always in the loop, right?
- 9:40SPEAKER_01: Because the human still needs to be part of part of the entire process.
- 9:43SPEAKER_01: But now you're thinking about the human as the reviewer and not the author, right?
- 9:48SPEAKER_01: Not having to do the administrative tasks.
- 9:50SPEAKER_01: But now you're looking at the agentic AI as somebody who is going to be the collaborator and working in the background so that you can continue to before, you know, continue to be doing the higher skilled work.
- 10:02SPEAKER_01: And so for a physical therapist, if you're going to ask me about in physical therapy, an AI agent could analyze a patient's progress from past sessions, maybe before the next visit, flag something that could be considered to be at risk, maybe not meeting a mobility goal.
- 10:23SPEAKER_01: The agent could maybe suggest alternative exercises, maybe draft a communication to the patient's family to provide some update or some additional information about the type of exercises or any kind of additional directions that need to be happened on the on their loved ones' progress.
- 10:41SPEAKER_01: GenTic AI could focus on analyzing an image of a wound and measure it, comparing it to prior images.
- 10:51SPEAKER_01: And if a wound isn't healing as expected, providing uh flagging it, triaging it.
- 10:56SPEAKER_01: But the human is always going to be in the loop.
- 10:59SPEAKER_01: We're using these tools as collaborators, uh, but it really is meant to be how do we use these tools to pre-populate necessary documentation?
- 11:09SPEAKER_01: Um, all of these, thinking about these agentic AI tools as collaborators for what I would almost call the mundane tasks, because we want to use our highly skilled staff for the highly skilled work.
- 11:25SPEAKER_01: Then also have their work be augmented by a powerful assistant in the background, right?
- 11:29SPEAKER_01: That can do the administrative work, that can do the heavy lifting of the data analysis, so that then we can focus on taking care of the patient and then being able to analyze the information and look at the analysis and make a determination rather than trying to sift through all the information and trying to bring the information together.
- 11:46SPEAKER_00: Absolutely.
- 11:47SPEAKER_00: I'm glad you bring up the human piece and you touched a little bit of administration.
- 11:51SPEAKER_00: I'm going to go with this a slightly different way.
- 11:54SPEAKER_00: Um, so when you work across the Google Cloud Healthcare's portfolio, the reality is you're dealing with some of the largest and the most highly resourced systems in the world.
- 12:04SPEAKER_00: For example, Mayo Clinic, HiMark Health, organizations that have massive dedicated AI teams and enterprise infrastructure.
- 12:14SPEAKER_00: But we all know that a significant portion of specialty care in this country is delivered by organizations that are running really lean, like your small skilled nursing facilities or your independent outpatient therapy practices or wound care clinics that have, and some of them don't even have a dedicated IT staff member, but some of them may just have a single person that's focusing on that.
- 12:36SPEAKER_00: So, in your experience, when you think about AI integration for a lean specialty with no dedicated IT staff, does the approach have to change completely or is that where the magic of AI comes into play?
- 12:50SPEAKER_01: Start with a really clear problem.
- 12:53SPEAKER_01: Focus on workflows and build trust with your employees.
- 12:57SPEAKER_01: If we're talking about specialty providers or specialty organizations with limited resources, the key is to leverage platforms that have actually already done the heavy lifting of building and integrating AI.
- 13:11SPEAKER_01: And I think that embedding AI into these workflows can actually deliver the power and the benefits of technology without actually needing or requiring a single dedicated IT member of a clinic.
- 13:28SPEAKER_01: And I think that's really where AI becomes a feature of a software that they already use every day.
- 13:34SPEAKER_01: AI is becoming infused into everything that is being done now in the healthcare ecosystem.
- 13:40SPEAKER_01: And so it's now becoming accessible to everyone.
- 13:43SPEAKER_01: And it's also becoming incumbent upon all the technology providers across the world to make it not only accessible, but also understandable by everyone so that it's not just for the IT providers or just for the tech people.
- 13:57SPEAKER_01: You'll hear me say a lot that now we are all tech people.
- 14:00SPEAKER_01: You're a tech person, I'm a tech person.
- 14:02SPEAKER_01: We are now all technology people.
- 14:04SPEAKER_01: And so it's also very important for technology providers to provide these technology tools in a way that is understandable not just by the IT department, but by everybody.
- 14:17SPEAKER_00: Yeah, totally.
- 14:18SPEAKER_00: And I'm um I I love that point that we're all IT specialists now, or we're all leveraging technology that's going to help us do our roles better.
- 14:28SPEAKER_00: Um, taking it in a slightly different, because this is actually something that's the road that I'm working walking right now, right?
- 14:34SPEAKER_00: As uh we decide around Martech tools that we're looking to power our organization.
- 14:39SPEAKER_00: One of the big questions that keeps coming up is ROI.
- 14:41SPEAKER_00: And one of the things I respect so much about you, Shwed, is that you just tell it like it is.
- 14:46SPEAKER_00: So, for example, one of the things I really enjoyed in our prep conversations is you've just been so direct about some of the things that a lot of the vendors just won't say.
- 14:55SPEAKER_00: Um, and your point there is that there's a big gap between running an AI pilot and really realizing enterprise value.
- 15:04SPEAKER_00: And in support of that specific point, I mean, there's a lot of you know studies out there.
- 15:08SPEAKER_00: There's one I think Deloitte did, uh, was it last year, something about the fact that while AI adoption in healthcare is definitely accelerating, they were seeing fewer than about a third of the organization really sort of reporting that AI had any kind of measurable ROI impact at scale.
- 15:26SPEAKER_00: So I'm kind of curious, at your specific vantage point, because you've worked with big and small, what are healthcare organizations actually seeing when they think about ROI from AI?
- 15:37SPEAKER_00: And what are they doing that others aren't and might need to be doing?
- 15:41SPEAKER_01: Uh so I appreciate your your comment about saying it as it is.
- 15:46SPEAKER_01: I think that's the only way I know how to communicate.
- 15:49SPEAKER_01: Uh, and I think, but I also think it it's important because it's it's this the these tools can be very um intimidating.
- 16:00SPEAKER_01: And I think we need to make them approachable.
- 16:02SPEAKER_01: And to make them approachable, we have to be very clear of what it can and what it cannot do.
- 16:08SPEAKER_01: And what can we do in the near term, what can we do in the medium term, and what can we do in the long term.
- 16:13SPEAKER_01: And I can in order to achieve that, you really, really have to start with the why.
- 16:18SPEAKER_01: So don't don't chase the shiny objects, right?
- 16:22SPEAKER_01: Uh, I often see when I'm sitting at the board level or with um executive, you know, excoms with CEOs and their direct reports, they're asking, you know, I need a, you know, I need a hammer and a nail.
- 16:33SPEAKER_01: I need to be, you know, this is the no, this is not what you mean.
- 16:36SPEAKER_01: You're not asking for a hammer and a nail, Francesca.
- 16:39SPEAKER_01: They're asking for a hole in the wall, right?
- 16:40SPEAKER_01: So, really, what is the why?
- 16:42SPEAKER_01: What are you trying to get to?
- 16:44SPEAKER_01: Um, particularly, I've seen in the last 18 months, uh, boards are saying, you know, we need to implement AI into our company, but why?
- 16:52SPEAKER_01: Why, right?
- 16:53SPEAKER_01: Do you really need it or do you just need to figure out how to get to this outcome?
- 16:57SPEAKER_01: Uh, so I think getting really clear about why, why you're why why are you going after this?
- 17:03SPEAKER_01: Identifying a very clear, specific, high impact clinical or operational problem, um, and then understanding if AI is a solution for that, right?
- 17:14SPEAKER_01: Um it could be, it not always is, right?
- 17:19SPEAKER_01: Um, and then I would say focusing on the last mile, right?
- 17:23SPEAKER_01: Understanding that AI, an AI model is only as good as the integration into your workflow.
- 17:29SPEAKER_01: So when you're designing solutions, think about who's going to use it.
- 17:34SPEAKER_01: If we're talking about in a practice, are we talking about talking to the folks who are working in the practice or the clinicians or the nurses or the physical therapists, exact, you know, et cetera?
- 17:44SPEAKER_01: Are we talking about the researchers who are on the bench?
- 17:47SPEAKER_01: All of this, right?
- 17:48SPEAKER_01: And to make sure that these things are intuitive and they're helpful and they get used.
- 17:54SPEAKER_01: Um, and then you have to build for scale from day one.
- 17:57SPEAKER_01: These can't be pilots and sit as these kind of point, you know, proof of concepts.
- 18:01SPEAKER_01: Um, we have to make sure that we choose partners that can and platforms that can grow as you are testing these things out.
- 18:09SPEAKER_01: And I think that's why um thinking about I mean, frankly, I will say thinking about cloud is so critical because what organizations should and can start small, but once you approve value, that scale is very important.
- 18:25SPEAKER_01: And you can do scale without having to do some massive upfront investment.
- 18:30SPEAKER_01: You can start small and then scale as you need, right?
- 18:33SPEAKER_01: And cloud can provide um that kind of flexibility.
- 18:37SPEAKER_01: Um, and then measure, right?
- 18:39SPEAKER_01: You need to know um what you're measuring, right?
- 18:43SPEAKER_01: The number of AI models is infinite.
- 18:45SPEAKER_01: So you have to know what you're trying to measure uh in order to understand what is the impact and what is the ROI that you're trying to measure.
- 18:53SPEAKER_01: Coming back to your original question, uh what is it, what what matters the most?
- 18:57SPEAKER_01: Is it your employee satisfaction?
- 18:59SPEAKER_01: Is it the outcomes that you're trying to, the uh the outcomes or the uh satisfaction of your patients?
- 19:04SPEAKER_01: Is it financial performance?
- 19:05SPEAKER_01: Is it speed or you know, some sort of operational efficiency?
- 19:09SPEAKER_01: So really measuring what matters the most has uh uh really makes for a very meaningful impact for these types of ROI uh projects for the use of AI in these uh is particularly in the healthcare and life science ecosystem.
- 19:23SPEAKER_00: I love that point about you know, really digging into the what are you trying to solve, right?
- 19:29SPEAKER_00: I think that's just it's it's so important to really look at workflow, diagnose gaps in the workflow.
- 19:35SPEAKER_00: And so I'm gonna throw one at you because I think this is a common problem.
- 19:39SPEAKER_00: Um, you know, moving from this whole notion that practitioners just despise this notion of documentation, right?
- 19:47SPEAKER_00: And it's the most consistent complaint when we do any kind of survey with any clinician across any specialty.
- 19:54SPEAKER_00: One of the biggest things that uh comes up is this documentation burden, irrespective of whether they're physical therapists, occupational therapists, wound care nurses, it's universal.
- 20:04SPEAKER_00: And the thing that is so uh challenging is that it's it's something that really follows them home, is sort of the quote that we hear a lot.
- 20:12SPEAKER_00: Um, there was a quote that I read in a recent report that said something like two hours of administrative work goes home with that doctor every night to every, and it's sort of like in proportion to one hour of patient care.
- 20:24SPEAKER_00: So think of that.
- 20:25SPEAKER_00: Two hours you're focused on administrative work for every one hour of patient care.
- 20:29SPEAKER_00: That's a lot of time that has been taken out of really what clinicians got into the business of doing, which is serving the patient.
- 20:36SPEAKER_00: Um, and you've talked about something which is really of interest, which I think starts to get to how do we chip away at that problem?
- 20:43SPEAKER_00: You talk about AI that can actually walk into a visit with intake that's already done, summaries that are prepared, follow-up actions already queued.
- 20:54SPEAKER_00: I mean, wow.
- 20:55SPEAKER_00: If an AI can do that and help prepare for a visit before it starts, that sounds super compelling.
- 21:02SPEAKER_00: But here's the catch, right?
- 21:03SPEAKER_00: With your experience with EHRs, what actually has to be true in the EHR for that to actually work?
- 21:10SPEAKER_01: The key is to make this a reality is for having the workflow with the I'm going to loosely call it the employees or the individuals who are in this, and having a modern record-keeping infrastructure all working together, right?
- 21:34SPEAKER_01: And that's when the technology can work at its finest.
- 21:37SPEAKER_01: And I will say, right, this this data, this stat that you're that you are um quoting, I will say it is likely very, very accurate because I have uh recently had um a dinner canceled on me uh because of a friend who is an OBGYN who had to cancel dinner with me uh because she had to stay home.
- 22:00SPEAKER_01: To finish some of the documentation and paperwork that she had to do.
- 22:04SPEAKER_01: She wasn't delivering a new baby.
- 22:08SPEAKER_01: It was the pajama time, right?
- 22:10SPEAKER_01: It was the pajama time documentation.
- 22:12SPEAKER_01: And so it was certainly absolutely true.
- 22:16SPEAKER_01: But for an angel to prepare, you know, for a visit, for it needs to actually read, you know, it has to be able to understand the data from a patient's journey, their past visits, the medications, the lab results, the information from the patient reported, um, everything that a clinician or a nurse would want to understand, right?
- 22:42SPEAKER_01: And that would require two things really, really robust interoperability, and the second is data fluidity, which I think is structured and standardized in a way that AI could very easily interpret.
- 23:01SPEAKER_01: We're getting closer.
- 23:03SPEAKER_01: And so once you have that foundation, I think that's when you can't, that's when AI is going to be able to figure out, right, if the blood pressure is trending up, or you know, if you need to put something in in the queue for a nurse to review, or whatever it might be, right?
- 23:19SPEAKER_01: And I think that is really when you're going to get to a point that rather than you're spending, you know, you could have a meaningful conversation with your patient rather than burying your head in the computer screen for that first 10 minutes.
- 23:33SPEAKER_01: And I think that's going to be the difference.
- 23:35SPEAKER_00: I'm so glad that you made that point about accuracy, because really, when we think about documentation from a records perspective as it aligns to reimbursement, it's really the key to getting paid.
- 23:49SPEAKER_00: I mean, and that's important for things like wound measurements or functional outcome scores or even just visit justification.
- 23:55SPEAKER_00: And we can't afford errors because if we have errors, that creates a compliance risk.
- 24:01SPEAKER_00: It also affects cash flow.
- 24:02SPEAKER_00: And in specialty care environments like the ones that NetHealth tends to serve, margins are already so thin that that difference really matters.
- 24:13SPEAKER_00: So, Schweta, where does clinical agentic AI actually move the needle most on accuracy and compliance?
- 24:20SPEAKER_01: So that is uh the North Star, I would say, in healthcare and healthcare AI, uh, moving from predictive or reactive to predictive to ultimately what I would say proactive.
- 24:33SPEAKER_01: And if you could look around what I keep saying is like the proverbial corners, so you could almost intervene before somebody's condition deteriorates or before what I say, right?
- 24:48SPEAKER_01: You know, um well care, not sick care, uh, before people have their uh condition or their symptoms actually showing up.
- 24:58SPEAKER_01: So this is how when you need AI to access longitudinal information beyond the four walls of a clinic.
- 25:08SPEAKER_01: This is you know, outside patient-reported information between visits.
- 25:13SPEAKER_01: So this is when you can have data from devices, monitoring devices, patient-reported outcomes.
- 25:20SPEAKER_01: This is where social determinants of health data comes in and the information from your records, right?
- 25:26SPEAKER_01: Which is foundational, but not enough.
- 25:29SPEAKER_01: This is where I think this, this is when you're going to have the real comprehensive view of individuals.
- 25:35SPEAKER_01: And then I think that's when you have the ability to take advantage, and an AI model can really identify the subtle patterns that humans can miss.
- 25:45SPEAKER_01: But when you have that information and collate that information, I think that's when you have the opportunity for the human or the expert to then actually be able to look at all this information and make determinations that maybe we otherwise would have missed because we would have just had that one data point to work off of, that one 15-minute interaction.
- 26:08SPEAKER_01: And so I think that's really when we can truly start to bend that curve, improve the outcomes.
- 26:15SPEAKER_01: I think we can have all the benefits to help some of the most vulnerable patient populations.
- 26:20SPEAKER_00: Let's take a step into the future, staying on the same threads of what you've just said, right?
- 26:25SPEAKER_00: And um, I'd say future, I should probably correct myself, rather at the speed at which AI is progressing.
- 26:32SPEAKER_00: Let's, it's probably the the now.
- 26:34SPEAKER_00: Um, but one of the key significant shifts, shifts that AI is helping make possible is us moving from that system that records what happened, like you said earlier, into one that anticipates what's coming next, right?
- 26:46SPEAKER_00: And then you actually gave some examples earlier where you said things like this could mean for a wound care patient, for example, that's at risk of hospitalization before deterioration starts to be visible in a chart, or catching a therapy patient, right, that's trending towards a plateau before that becomes um, you know, some sort of a different conversation.
- 27:07SPEAKER_00: So just staying on that specific piece there and also tying to what you just mentioned uh a little uh while ago about agentic clinical AI, what does it actually take for proactive patient journey management to work, both from this clinical and technical standpoint?
- 27:23SPEAKER_01: My advice is always to start with the low hanging fruit.
- 27:29SPEAKER_01: The tasks that take up the most time and that causes the most frustration are probably the things that you should go after, right?
- 27:39SPEAKER_01: So in specialty providers, I think that some of that is documentation.
- 27:46SPEAKER_01: I think that's some of the in when in my experiences across the healthcare and life science ecosystem, I think it's automated documentation is one of the highest ROI use cases for AI by far.
- 28:02SPEAKER_01: Using AI to automate the creation of notes, of lab data, of what have you, but the automation of you know, this collating of information and getting insights, you can give back hours in a day.
- 28:19SPEAKER_01: I think that there's opportunity for intelligent scheduling, and anywhere that there's prioritization and task management, and an AI agent can act as a virtual assistant.
- 28:30SPEAKER_01: And in this case, we're talking about for you know, for nurses, for clinicians, for those who are working in specialty management.
- 28:37SPEAKER_01: This can help you prioritize your day, triage your day, and ensure that the most urgent tasks are always at the top of your list.
- 28:46SPEAKER_01: And think about that, that means you're talking about this in a uh professional context, but don't you also wish, Francesca, that you could do this in a personal context as well, right?
- 28:54SPEAKER_01: That you know, you could prioritize the things that actually need your attention the most.
- 29:00SPEAKER_01: But this is the most valuable, I think, across the board.
- 29:03SPEAKER_01: This is not specific just to our industry, but we're seeing this as you know, scheduling, task management, and documentation, lowest hanging fruit, highest ROI.
- 29:16SPEAKER_01: And um these are things that are these are capabilities that are proving high ROI across multiple, multiple sectors um already.
- 29:29SPEAKER_01: And if you take the time to see uh various reports that are coming out, you'll start to see that this is not just a, oh, this is a great idea.
- 29:38SPEAKER_01: There is already organizations in our industry and outside of our industry that are already showing statistically significant ROI from this.
- 29:47SPEAKER_00: That's fantastic.
- 29:48SPEAKER_00: Um I and I love the fact that you brought up specialization because one of the things that we bump up against quite a lot with our um providers, and of course, there's tons of major EHR platforms out there that are moving really fast on the AI front.
- 30:05SPEAKER_00: I think EPIX uh recently on their um their annual conference, they quoted something like 150 features that are in development for 2026.
- 30:16SPEAKER_00: And those are all native agent capabilities that are built right into those platforms.
- 30:20SPEAKER_00: And like you said, for a large healthcare system, that's really compelling.
- 30:25SPEAKER_00: But where it starts to get kind of tricky is when you start to talk about nuanced specializations like a wound care or a rehab therapy or occupational health kind of workflow, those are not just generic medicine workflows.
- 30:40SPEAKER_00: And that's where those documentation requirements or outcome measures or even that reimbursement logic that we were talking about earlier to get paid, that's all fundamentally different.
- 30:50SPEAKER_00: And what we always tend to think is that a platform, in fact, it's really the reason why NetHealth is here, the origin of it, is because a platform that's built for everything is really by definition not really built to help these types of domains.
- 31:05SPEAKER_00: So when it comes to specialty care in AI, what does deep and really specific workflow make possible that a generalized platform just simply cannot deliver?
- 31:16SPEAKER_01: Even the most sophisticated AI model can be not can be considered to be not useful unless you have full transparency into all of this work, right?
- 31:30SPEAKER_01: And the workflows.
- 31:31SPEAKER_01: I think that exactly into the work, into the workflows, into uh and and being able to understand what's under the hood.
- 31:40SPEAKER_01: If you remember earlier in our conversation, we were talking about that we are all technology people.
- 31:45SPEAKER_01: And what that means is we all need to understand, you know, this can't just be a black box.
- 31:50SPEAKER_01: This also needs this needs to be a glass box.
- 31:52SPEAKER_01: And what that means is not only is it, you know, do the technology providers provide the information about what, how, why did the algorithm provide this as the output, but we on the uh, you know, in the clinic need to understand what what was this output and why did this output, you know, why was this output?
- 32:10SPEAKER_01: So those who are working in the clinics always have to be in control.
- 32:15SPEAKER_01: We need to look at AI as the advisor, not the black box that's making the decision.
- 32:19SPEAKER_01: It's the triage, it's the collaborator.
- 32:22SPEAKER_01: Um, and so I think part of that is um, you know, we're we're we are still talking about patients, we are still talking about a regulated industry.
- 32:32SPEAKER_01: And so while I have so much trust and um so much trust and hope for this application of technology in in our ecosystem, we still need to work with low-risk use cases, start small and then and then scale.
- 32:52SPEAKER_01: And I will always say, right, rigorous validation and governance is always going to be required.
- 32:57SPEAKER_01: You have to have a robust process for validating all of these AI models and modic and monitor them over time.
- 33:03SPEAKER_01: So I'm going to always, of course, put a plug on, you know, at Google we have you know a set of responsible AI principles that guide everything that you know we do and we work closely with partners to develop the governance frameworks.
- 33:15SPEAKER_01: But I and I know I'm digressing a little bit, Francesca, but you know, building trust in AI is not a one-time event, right?
- 33:24SPEAKER_01: And part of working with you know the the uh the folks that you serve, right?
- 33:29SPEAKER_01: It's an ongoing process of communication and collaboration and continuous improvement that I think that helps build part of that uh part of that care.
- 33:39SPEAKER_00: You know, the fact that you brought up trust is actually key because something that sort of plagues the AI conversation is also this notion of confusion around, hey, AI is really going to come in, we're gonna have all these efficiency gains, and then we're gonna use it as a justification for reducing headcount rather than really focusing on improving care.
- 34:00SPEAKER_00: And I don't know that that's necessarily a hypothetical because we've seen so many healthcare systems, candidly, tons of SaaS organizations as well, also citing that AI productivity is really a factor for staffing decisions.
- 34:15SPEAKER_00: So again, coming back to some of the fact, you know, the the notion that some are big hospital organizations, but a good chunk are these lean speciality practices.
- 34:24SPEAKER_00: How should they think about deploying AI in a way that really improves care rather than just reducing headcount?
- 34:31SPEAKER_01: If I could leave you and the audience with one thing, it would actually be don't wait, right?
- 34:39SPEAKER_01: The future of our ecosystem and life science ecosystem is actually being built today with AI at the center of it.
- 34:48SPEAKER_01: A few years ago, I would have said, you know, and I coming from the healthcare system, I would say, right, technology was, you know, considered to be an innovation project, right?
- 34:59SPEAKER_01: That if you had a little bit of budget dollars, you would experiment with some technology tools, but really your budget dollars were going to the core of your of your work.
- 35:07SPEAKER_01: Uh now I would say technology is a foundational part of how the healthcare ecosystem is going to work and is working.
- 35:15SPEAKER_01: So don't wait.
- 35:17SPEAKER_01: You don't need a massive budget.
- 35:19SPEAKER_01: I'm going to say the same thing.
- 35:20SPEAKER_01: You need a clear vision.
- 35:22SPEAKER_01: You have to be a willingness to experiment in a very concise and clear way in a commitment to put your employees at the center of everything of how you're going to work, right?
- 35:36SPEAKER_01: Your employees, and then of course, therefore, your the the patients and the individuals you serve.
- 35:41SPEAKER_01: So I would say today's today's that we're at the end of the week.
- 35:45SPEAKER_01: On Monday, I would encourage anyone who's listening or the you know, the next week have a conversation with your team, right?
- 35:52SPEAKER_01: If you have a magic wand, what would be the one thing you would change with your workflow?
- 35:57SPEAKER_01: And I guarantee that the answers that you give, that you get would give you a roadmap of where to start with AI.
- 36:04SPEAKER_01: So start small, think big, and move fast.
- 36:07SPEAKER_00: That would be how I would say.
- 36:09SPEAKER_00: I absolutely love that.
- 36:10SPEAKER_00: And on that note, I really think this is an appropriate place for us to wrap our conversation.
- 36:16SPEAKER_00: Um, we're gonna let Shweta off the hook before we go into our top three takeaways from today.
- 36:21SPEAKER_00: Shweta, thank you so much for joining us.
- 36:23SPEAKER_00: This was a genuinely illuminating conversation, and we are so looking forward to seeing you at NetHealth next in April.
- 36:30SPEAKER_00: Thanks again, Shweta.
- 36:32SPEAKER_00: Thank you so much for having me.
- 36:38SPEAKER_00: So we covered a lot of ground in today's chat.
- 36:41SPEAKER_00: So pulling out just three takeaways was actually really tricky.
- 36:45SPEAKER_00: But here's what stood out to me the most.
- 36:47SPEAKER_00: First, the EHR's original promise broke down because documentation was never really designed around the clinician or the patient.
- 36:55SPEAKER_00: AI has the potential to rebuild that workflow entirely from pre-visit prep to post-visit follow-up.
- 37:02SPEAKER_00: So the record better serves the people that are actually creating it and the people that it's created for.
- 37:08SPEAKER_00: Such a powerful point.
- 37:09SPEAKER_00: Second, agentic AI is the evolution worth watching.
- 37:13SPEAKER_00: We're moving past tools that just listen and transcribe into systems that are starting to anticipate, prepare, and act proactively while keeping clinicians all in the driver's seat.
- 37:25SPEAKER_00: For speciality care in particular, that distinction between ambient documentation and goal-directed AI agents carries real consequences for how organizations plan, spend, and ultimately deliver their care.
- 37:39SPEAKER_00: And finally, trust is paramount to any implementation strategy.
- 37:44SPEAKER_00: Adoption rates mean nothing if the clinicians using these tools don't truly believe in them.
- 37:49SPEAKER_00: Governance frameworks, workflow specificity, and transparency about how AI-generated outputs are derived is what separates a successful deployment from one that really quickly gets abandoned.
- 38:01SPEAKER_00: And that should do it for this episode.
- 38:03SPEAKER_00: As always, I'd like to thank all of our listeners for tuning in.
- 38:06SPEAKER_00: If you're joining us on Spotify or YouTube, we'd love to hear your thoughts in the comments.
- 38:11SPEAKER_00: And don't forget to leave us a rating while you're there.
- 38:14SPEAKER_00: Once again, I'm Francesca Ferrab, head of marketing at NetHealth, and I'll see you next time.