Q: Hi Michael. Welcome to the Big Unlock podcast. Very excited to have you here. The healthcare audience is super excited to listen to our podcast. I’m Rohit, the Managing Partner and CEO at BigRio and Damo Consulting. With that, I’d like to request you to introduce yourself, Michael. 

Michael: Yeah, I’m excited to be here too. I’m Michael Marchant, Director for Digital Applications at Sutter Health. I’ve been here almost a year now, focusing on core interoperability, enterprise data integration, and also running our Community Connect program with about 250 practices.

I’ve been in interoperability for a long time. I’ve participated on the Epic Care Everywhere Governing Council, I’m on the Carequality Advisory Council, and I serve on the eHealth Exchange Board. So, interoperability and health technology are core to what I do every day—making decisions and planning for how we get people the data they need to make decisions around their healthcare. Excited to be here.

Q: That’s great. Michael, would you like to share your journey in healthcare? What got you started? What roles have you played, and what do you find interesting about this field? 

Michael: It’s a long and interesting story. I went to college in Virginia at Old Dominion University, studying Human Resource Management. I started as an internal auditor. My father worked for a small software company called Comcare, and when I moved to California, they were looking for someone to install patient financial software—AP, GL, payroll, personnel.

I had some technical background but no real experience in software implementation. They were willing to train me, and I got on board. My first foray into interoperability was doing payroll installs—sending data from point A to point B to ensure employees got paid. That was my first critical experience with data exchange.

I moved from payroll into accounts payable and general ledger—again, lots of data movement and system frameworks. Comcare had a partnership with DataGate, which had an interface engine, and that’s where I started programming. That kicked off my career in health IT.

After having kids, Southern California became expensive, and my wife, despite her advanced degree, wanted to stay home with the kids. So, we moved to Sacramento, and I started with Sutter. I was part of the Y2K program with MedSeries4, working on integration across sites to ensure interfaces were Y2K-compliant—moving from two-digit to four-digit years and ensuring backward compatibility.

At Sutter, I moved up from analyst and project manager to director. Sutter Connect brought me over to run what we then called the Independent Practice Services program—it wasn’t yet Community Connect. We deployed GE’s product (GE had acquired IDX) to about 50 practices and 100 physicians before I moved on.

After that, I worked with the state contractor for the EHR Incentive Program. ACS Xerox partnered with California’s Department of Healthcare Services to build the state’s enrollment portal for incentive funds. I helped with training, enrollment, and deployment. By the end, we had paid out about $500 million in incentives.

Then I went to UC Davis, where I was Director of the Interop and Innovation Team. That’s when I got deeply involved in the industry—Carequality, HL7’s Interop Committee, and the blockchain task force. I still believe blockchain has potential, even if people aren’t quite comfortable with it yet.

I spent about 10 years at UC Davis and recently returned to Sutter, leading two teams. For me, technology is in the background, serving a purpose—automating, simplifying, accelerating. It’s about getting the right information to the right people so they can make the best decisions about care.

That’s what drives me: improving processes, delivering value to both internal and external stakeholders—our physicians, clinicians, and most importantly, the patients. This work creates better care, and I try to reinforce that with our teams. Even when you’re buried in the weeds of programming or writing specs, I try to help everyone see the forest through the trees.

Q: That’s awesome. And Michael, could you talk to us at a high level about Sutter as an organization—patients being served, locations, hospitals, physicians, and overall size? 

Michael: For sure. Sutter Health is a large organization in Northern California. The Sutter footprint goes from about Sacramento, or a little north and east of Sacramento, down to Santa Barbara. We recently acquired the Sansum Clinics in the Santa Barbara area. So, from north of Sacramento to Santa Barbara, you’ll find Sutter entities and coverage.

We have around 25 to 27 hospitals, a number of ambulatory surgery centers, and we’re continuing to grow in the ASC space. We’re building a couple of cancer centers and a few other joint ventures. We really serve the population of Northern California—probably 2 to 3 million covered lives from a value-based care perspective, and hundreds of thousands of visits. We have a strong online presence as well as a physical presence in many Northern California communities, always aiming to deliver the best care.

Sutter has won a lot of awards—I don’t have my slide deck in front of me, but if I were doing a PowerPoint, I’d have the slide with all the awards.

Q: This high-level information is awesome, thank you. It’s good to know—it’s a pretty large footprint. With that, I’d like to segue into the first question. We were talking earlier about the CMS and ONC RFI that’s out right now and the TEFCA implementations you’ve seen to date. How do you think about this emerging process? 

Michael: Yeah, I’ll start with TEFCA and then go to the RFI. The general idea from ONC, now under CMS, is to expand the capabilities of our national networks.

Historically, eHealth Exchange, CommonWell, and Carequality were the three national networks enabling organizations to do nationwide data exchange—mostly CCD documents—on a request-response basis, securing the connections and delivering data to the requesting organizations.

We’ve grown that network significantly. In a recent committee meeting, we learned that 1.1 billion documents were exchanged in April alone between participating organizations. The structure of those networks is treatment-based exchange—organizations treating patients and needing information to assess and understand the patient’s history to provide the best real-time care.

But we kind of stalled there. We were stuck at treatment-only. Many organizations that needed access to information couldn’t participate. During COVID, public health access became paramount, and CMS allowed some exceptions under emergency orders, which enabled that information exchange. But for broader, public health, and operational exchanges—such as care gap closures in value-based populations—those aren’t widely supported on treatment-based networks.

TEFCA’s promise is to expand this. Many healthcare ecosystem organizations—long-term care, SNFs, nursing homes, and community providers—weren’t eligible under the EHR Incentive Program and thus aren’t participants in the current networks. Roughly 30% of healthcare data holders aren’t participating in current treatment-based exchanges. TEFCA aims to bring them in and broaden the scope of who can participate.

It’s not just about treatment. The next phases involve operational exchange, payment exchange, public health, and research. The idea is to build a network that supports all of these use cases and welcomes organizations that couldn’t participate before—either because they’re not involved in treatment or they’re not HIPAA-covered entities.

CMS is taking the lead. There are now 8 to 10 QHINs (Qualified Health Information Networks). Most EHR vendors, Surescripts, eHealth Exchange, KONZA, and others are participating. The RFI seeks broad input on how to support this ecosystem: improving provider adoption and usability of digital tools, strengthening data exchange, ensuring access to comprehensive patient data, and advancing value-based care through technology.

It’s about identifying data needs, reducing administrative burdens, enabling patient and caregiver access to digital tools, managing identity, ensuring equity in digital health access, and setting future-proof certification criteria for participation.

I encourage everyone to comment and provide feedback to CMS based on their experiences—so these standards and frameworks are implemented correctly. That participation is vital.

Q: That’s awesome to know. Now, moving forward, Michael—on digital health tools, AI, and of course now GenAI and all the new LLMs coming into play.
Could you talk with us about the risk profiles and risk appetite for health systems, and share some examples you’ve seen at Sutter or at UC Davis? 

Michael: Yeah, for sure. AI is used for a large number of things—NLP, machine learning, RPA—those have been in place for a while, and we’ve been leveraging them. Now we’ve wrapped them under the umbrella of AI, but the buzz right now is around GenAI, especially ambient voice.

The idea is a physician in the room with a patient, having a conversation while an ambient agent listens in the background. It generates a note at the end of the encounter, allowing the physician to focus on the patient, not the computer. It makes the interaction more personal. The agent formats the note with relevant information while excluding casual conversation.

This technology has helped physicians get off the keyboard and return to practicing medicine. Satisfaction scores are high. It’s reduced pajama time—charting after hours—for those using it. We’re working to expand adoption across the board.

We’ve seen similar value in Epic. There’s an inbox agent that can read patient messages, prioritize them, and even draft responses. A physician managing a thousand patients receives a lot of messages. The AI agent helps flag urgent ones and draft thoughtful replies. In some cases, the AI-written replies are more empathetic than the rushed responses doctors might send. It’s all reviewed and approved by the physician, of course, but the assistive value is significant.

In radiology, AI helps prioritize breast imaging. It flags anything abnormal and puts it at the top of the radiologist’s worklist instead of first-in, first-out. That ensures patients with serious concerns are seen sooner. Given the shortage of radiologists, this makes a big difference.

We also implemented a tool for diabetic retinopathy screening. It takes images of the eye and scores them. An untrained person can operate the device, which helps us scale the service. Previously, there was a long wait for tests—now we can screen thousands of patients and prioritize those needing care.

So across the board, these tools are assistive. They don’t replace clinicians but help deliver faster, more targeted care.

Q: That’s great to know. So do you think, Michael, that at any point, any of this AI — in the examples you might be working on or seeing — is going to be autonomous? 

Michael: I think there are probably some places where autonomous activity can occur in low risk — and that’s where you move to back office, right? A lot of the patient-facing care applications are going to be assistive or augmented rather than delivering the final result.

Things in the administrative area — like claims and claims status — a lot of that is logging into portals, sending messages, or even sending a fax. Those can be automated. Responses can be automated. You can OCR faxes and apply some RPA or ML rules to those transactions and execute commands. It happens in a repeatable, machine-like way.

But when you’re talking about a patient having a procedure or getting a test or result, the AI can help look at larger datasets. Say you have a patient with a specific condition. You can use an LLM to say, “I have Rohit, who has this condition. What are the outcomes for others with similar conditions? What medications did they take?” You help frame a care plan based on data from 100,000 patients instead of anecdotal information or outdated knowledge. So you’ll see more research-integrated LLM data that allows physicians to target patients with similar profiles — but it’ll still be assistive and suggestive, not making final decisions.

Q: That’s a good way to think of it. The key thing here is the risk factor. Lowest risk things you might automate; high risk, you lean towards assistance. 

Michael: Yeah, I’ll give you an easy example here. I’m in IT, so I’m participating in a pilot with a blood pressure device.

What happens is I take my blood pressure, and the reading goes to Epic. Part of the AI in the background looks at the readings, and if mine is not what it should be, it sends me a text saying, “Hey, your reading was out of band. You might want to take a sip of water, relax for a second, and retake your blood pressure in 15 minutes.”

All of that is low risk and automated. Nobody has to review the score. It knows it’s out of band, and it recognizes there might be some environmental factors impacting the reading. So it automatically sends me a suggestion: “Hey, your blood pressure was high. Could you take another reading in 15 minutes after calming down or making sure nothing was off?”

That sort of automation is low risk. It reviews the data — so there’s some oversight — and it actually gives the patient comfort. It makes you think, “Oh wow, someone’s paying attention,” and that kind of outreach builds confidence. It’s a patient satisfier. It improves safety. There are a lot of benefits — and again, it’s low risk.

You know, I’ve done it in the office — taken a reading, it was a little high, and they’d say, “Okay, let’s take it again.” Two minutes later, it’s normal. So this is similar. And when you think about caring for a larger set of patients, we’re going to distribute that blood pressure cuff to thousands of people. We don’t need a ton of nurses looking at every result. We just need the rules in place: if something’s out of band, ask for another reading. If it’s out again, then alert a clinician.

Then, that physician or clinician might say, “Hey, you’ve had a couple high readings in a row. We may need to adjust your medication. We might need to bring you in. Or, this looks serious — please go to the ER.”

That’s where you apply the right resources in the right way. AI becomes a tool to help prioritize risk. For patients who are higher risk — and where there’s information you can act on quickly — a person can step in rather than relying solely on tech.

As we get more advanced, we’ll expand this kind of approach into other use cases beyond blood pressure to truly provide personalized medicine. Back when Bush talked about personalized medicine, it was tied to genetics and DNA. But I think personalized medicine is really about treating each patient as an individual and giving them feedback and support to manage their condition proactively — before there’s an adverse health event.

That’s what value-based care is about. We want to keep people out of the hospital and care for them before they get sick. These low-risk AI tools make a lot of sense for that. We just need to figure out a reimbursement and incentive model that encourages both patients and providers to adopt them — instead of waiting until people get sick and need hospital care, which, of course, is expensive for everyone.

Q: Of course. So Michael, moving on to another aspect of innovation—thinking about it in the framework of new announcements. Oracle has announced a completely new redesign of the EMR system, and we all know Epic is the big, heavily used system by health plans and health systems. How do you think about innovation in this scenario?

Michael: One of the things about Epic—and again, Epic is a significant vendor in the electronic health record space—they’re a market leader. They deliver great technology. I’ve been a customer of Epic since 1998, so I’m very familiar with them and their architecture.

I think one of the key aspects of modernizing electronic health records is abstracting the data, workflow, and technology. Epic, like most of the historical EHRs, was built first as a billing system, and then as a clinical documentation system. A lot of the workflows inside current EHRs, including Epic, are really tied to documenting the right things so that when we send the claim to the payer, it’s approved—not denied—because we’ve checked all the right boxes or coded it correctly.

Looking at Oracle’s new ambitions—they’re focused on an AI-native, cloud-native, redesigned user experience that really puts clinical workflow first. Instead of thinking about billing first, they’re thinking about the clinician experience first, and then reverse-engineering it to handle administrative tasks using AI and other tools.

Whether it’s Epic, Meditech, Cerner, Oracle, Allscripts, Greenway, Athena—you name it—the future is about enabling the right workflow for the right specialty and clinician. There should be personalization that allows them to work in a way that makes sense to them, while still giving access to all the necessary data.

If we can create data liquidity—which is what they’re aiming for with FHIR APIs and API-enablement on the back end—then we’re no longer chained to Epic or Oracle’s UI. If we can access the underlying data structures, we can build apps on top of that to create more fluid, user-based workflows.

We’ve gone through different UI development phases—widgets, phone apps, watch apps—and the future will be about building UIs that are intuitive for clinicians. They should meet clinical needs first, and then use AI and automation to handle the back end—like billing, compliance, and documentation. That way, we’re not robbing innovation on the front end just to serve billing on the back end.

You’ll prompt clinicians when needed—like, “Did you ask about smoking status?”—instead of forcing them to check endless boxes. Whether it’s Oracle or a smart startup, someone is going to leverage existing EHRs—like Epic’s infrastructure and data—but create a much better UI and clinical experience on top.

Replacing an EHR is incredibly expensive. So instead of ripping and replacing Epic, maybe you introduce a SMART-on-FHIR application for cardiology that works better than what Epic has. You get some market traction, then build apps for other specialties.

You’re not going best-of-breed across everything—just best-of-breed on UI—while still using your existing EHR’s infrastructure and data model to drive the new experience.

Q: That’s a great perspective on innovation, and I really like the term you used—data liquidity. It’s an awesome way to think about it. As we come toward the end of our podcast, Michael, I wanted to touch on one more topic—the California Data Exchange Framework, which you mentioned earlier. Tell us more about that and where you see it going.

Michael: Yeah, so California—like every state—thinks they’re unique. Governor Newsom signed the Data Exchange Framework legislation a couple of years ago. It’s essentially establishing a parallel data exchange within the state that allows a person living in California to access their longitudinal health record, including social services data and public health data.

They’ve created a TEFCA-like model with QHIOs—Qualified Health Information Organizations—that help facilitate participation. There are requirements around admit notifications, panel management, direct messaging, and data exchange for treatment and public health operations.

Right now, the Data Exchange Framework is a legal framework, but not a technology framework. They’re trying to be technology-agnostic and allow organizations to participate using whatever tools they currently have.

One challenge is the need for point-to-point connections. If I want to exchange data with, say, a housing agency or public health department in Sacramento, I have to figure that out directly with them. But the legal framework provides cover—a data use agreement that says the state has authorized the exchange, so we don’t need separate agreements.

Many groups involved—like housing or public health—haven’t historically participated in data exchange. So the questions are: What data can public health have? What can a housing agency or non-covered entity access? How do we ensure the right security profile? Who’s authorized to participate?

Today, to find out who’s signed the Data Exchange Framework, you have to visit a website and look them up on an Excel spreadsheet. That’s obviously not scalable. The future needs to include a real technical infrastructure to support this exchange—not just point-to-point connections, but a true network effect.

It’s similar to TEFCA but running in parallel. I’m working with BluePath Health, EMI Advisors, and others on this in California. I serve on the Technical Advisory Committee led by CDII to provide feedback and help define the standards and frameworks to enable this exchange.

It’s an exciting initiative. SB 660 is moving through committee and will help enforce participation—not just in signing agreements, but in actually exchanging data. That’ll give the framework some real teeth.

Of course, there are challenges—AB 352, for instance, which prevents abortion-related data from leaving the state. So there are lots of complexities. But California is serious about interoperability, governance, and regulation, and we’re all working to comply. Hopefully, this framework will help us get there.

Q: That’s awesome, Michael. So, any other parting thoughts for the listeners of the podcast from your side before we wrap up?

Michael: There’s a lot happening, and it can feel overwhelming. What I always tell people—like I tell my kids going off to college—is: get involved. Even if it’s something small, participate. Find people who are involved, ask questions, learn from them.

Many of the things I’ve learned came from conversations with really smart people. I participate in the Health Tech Talk Show with Lisa Bari and Kat McDavitt—they’re incredibly knowledgeable. Lisa was the former CEO at Civitas and now works at Innovaccer. Brendan Keeler also shares great content on LinkedIn and through his blog.

Join a committee. Help make sure your organization is ready. If you’re in the interoperability or health tech space and you’re not moving forward, you’re going to get left behind.

Engage, understand, and provide feedback—especially to things like RFIs. If you haven’t read them, plug them into your AI, get a summary, and start there. Use the tools we have.

And I always tell people: I’m an open book. If you have questions, feel free to connect with me on LinkedIn—we can have a great conversation. Thanks again for having me on the podcast, Rohit. Exciting times ahead for all of us.

———————— 

Subscribe to our podcast series at www.thebigunlock.com and write us at [email protected]   

Disclaimer: This Q&A has been derived from the podcast transcript and has been edited for readability and clarity.

Q: Hi Mike. Good to have you on the podcast today. Mike. I’m Rohit Mahajan, Managing Partner and CEO at BigRio and Damo Consulting. I’m very fortunate to carry on The Big Unlock podcast, which is the legacy of Paddy Padmanabhan, who was the founder of Damo Consulting. We are, I think, getting well over 170 podcasts at this point. I’m super excited to have you here today.

Mike: Hey, Rohit, great to be on. Thanks for inviting me. I am currently the Chief Transformation and Innovation Officer at United Church Homes. We are a nonprofit provider of senior living, with over a hundred different properties in 15 states and on two tribal nations. That includes about 75 affordable housing properties—owned and managed.

And then in the state of Ohio, where we’re headquartered, we have 10 owned and managed skilled nursing communities. We have life plan communities, and we also have independent living communities focused on the middle market.

We’ve definitely been growing—a very innovative organization that I’m proud to be part of. That includes growth into more services. We’re starting to do programs with CMS, like the GUIDE program. We also have a joint venture with a managed care payer called CareSource.

I really think that’s the future for many senior living providers—growing from being just housing providers to becoming health and wellness providers, with housing at the core.

Q: That’s amazing. So Mike, would you like to share with us your journey in healthcare? What got you started, how did you get to where you are, and what are some of the things you’re seeing for the future? 

Mike: Well, thanks for asking the question. When I started my career, I came down to the U.S. from Canada right out of grad school. I got into advertising, and with my first degree in economics, I became very compelled by the age wave.

I think there’s nothing more predictive—outside of maybe climate change—of future demand than the current age wave. It’s what I call the anomaly of the baby boom generation. This huge baby boom spike—we’re sitting here on July 14, 2025, and in 2027, we’re going to have the most people turning 80 in any year ever. Why? Because of what happened in 1947.

We know people needing healthcare services today are just a small fraction of what we’ll see in the future. So why not get in and plan for it?

Being a very naive Canadian at the time, I thought, “Maybe I should just get into the healthcare space—because that’s easy in the U.S., right?” Very straightforward! But it’s been compelling.

I started at AARP. They were a client of mine when I worked in advertising, and they’re the largest association for older people in the country. Then I moved into health IT with a role at Surescripts, the nation’s e-prescribing network, and later returned to AARP. So I got a taste of interoperability, health tech, and how we could apply that to aging.

With labor shortages and so many constraints, I started asking: how can technology fill the gap and support people at scale?

Through further experience, I realized clinical care isn’t the most important factor in the health and wellness of older people. It’s the non-clinical care and functional support that matter more.

So I’ve built expertise in social determinants of health—risk modeling and strategies to support aging at home. And importantly, that doesn’t just mean care and safety. I think it’s far more effective to support someone’s needs when there’s real meaning behind it.

One of my favorite sayings is: “Nobody takes their pills because they like how they taste.”

I’m a big fan of more relational care models, where we work hand in hand with people—focusing on their personal goals and motivations. And I hope that at United Church Homes, and in the future, that’s the kind of model I can help advance.

Q: That’s an amazing journey, Mike. Given your deep immersion and expertise in this area, would you like to talk about some growth strategies for senior living care?

Mike: I’ve had the privilege of working with United Church Homes for almost four years, and it’s helped me really get to know the nonprofit senior living industry. I’ve been incredibly impressed. As I like to say—we do the most with the least.

Especially in affordable housing. For about 30 years, United Church Homes and others in our space have been participating in a program that HUD (Housing and Urban Development) started called Service Coordination in Multifamily Housing. They provide funding for staff to be at those properties and conduct social determinant assessments on residents.

We help reduce risks by connecting residents to local community resources they may not know about. We also help qualify people for Medicaid and Medicare waivers. More importantly, we talk people down off a cliff.

What I mean is, when someone is going through a health or aging challenge for the first or second time—we’ve been through it hundreds of times. So we can tell people what’s normal, what’s not, and what might happen next.

That program has been amazingly effective in keeping people out of the hospital and skilled nursing. We have about 3,800 people in affordable housing—3,200 of them are on service coordination contracts. In the last 15 months, only 50 transitioned into skilled nursing, and 110 had an unplanned hospitalization. These are very low numbers—and they’re similar to others in our space.

So when I think about growth in senior living, I think about unbundling this service skillset and offering it as a standalone solution. It’s a plug-in for managed care, employer programs, and long-term care insurance. I think the future of senior living is transitioning from housing providers to health and wellness providers—with housing at the core.

I also think the industry needs to shift from centralized service delivery to a more decentralized model—because there’s an opportunity to support people in housing who could never afford or don’t want to move into a community. A hub-and-spoke model—I think we’ll see more of that in senior living.

Q: That’s very interesting, Mike. Just curious—typically at what age do people come into these kinds of facilities? What have you seen, and is that changing? 

Mike: Yeah, I mean, average age is going up—we’re getting them older and sicker. The lifetime value is going down. So the business strategy has to diversify toward more community-based models—serving people where they are.

The average time between knowing you need to move into a community and actually doing it can be 9 to 18 months. What happens in between? That’s where we need technology to help. Technology that tells us when someone may need assistance—and also helps capture that data.

Q: Right, of course. And you talked about social determinants of health—that caught my attention. How does that play into your transformation and innovation efforts? And what are some things you’re doing with machine learning or GenAI in this space? 

Mike: Social determinants impact about 70% of health outcomes. Clinical care is about 10%, genetics about 20%. But social determinants cover everything from food, transportation, and shelter to how people engage with their health—adherence to care plans and motivation.

That’s why the service coordination model is so important. It builds trust first. Then it designs care plans around personal motivations—like wanting to keep a dog, visit a garden every day, or attend a museum show. These are real examples.

We need to understand what impacts those motivations. I like to call it “micro social determinants.” The macro ones are where you live, education, resource access, etc. But the micro ones are things like: do you have a primary care doctor? Can you get to appointments? Do you have a reconciled med list? Can you follow your meds? What’s your functional status?

Because if you have three or more chronic conditions, you cost about 50% more. Add one functional limitation—and that jumps to 330%. That’s about 5% of patients consuming 25% of healthcare spending. They trip, fall, and go to the most expensive sites of care.

So home safety—clutter, lighting, cords—all matter. Caregiver presence and quality matters. Your own goals and motivations matter. These are social determinants.

Another big insight from my career—when I was at AARP, we did a study on health literacy. The doctor sees me, puts on a blood pressure cuff, and says “119 over 70-something.” I ask, “Is that good?” He says “Yeah,” but it means nothing to me.

People over 65 are the least health literate—but when they are literate, they’re the most adherent to care plans. So again, that supports a social determinants and relational care model—to reduce spending on the highest-cost patients.

That’s what I get excited about in data analytics. I don’t think we’ll ever get to a point where we have social determinant care pathways like clinical ones—like in cancer, where you try drug A, then chemo B.

Nonclinical care has so many variables—social, financial, environmental, motivational—but maybe we can get close.

Q: So in your journey, what have you been experimenting with or piloting in terms of new technologies like GenAI or AI in this space, or Internet of Things—devices to keep this patient population safe and risk-free?

Mike: Appreciate that. Yeah—so first, machine learning is my top priority for innovation right now. Just like I said before: how can we take all the data we’re collecting on social determinants, the referrals we make to local community programs, and the efficacy of those programs?

We often know the best home health providers in the area. Just last month, a woman had bedbugs, and we knew Catholic Charities has a furniture bank that helps with new furniture. So how can we take both the referral and the result information, the social determinant data, and model it into efficacy?

Because that’s going to be our pathway into managed care programs. The biggest challenge for our industry is that we can’t take risks within managed care yet—our data game isn’t strong enough. When we combine clinical care and nonclinical care, it’s like baking a cake and trying to take the eggs back out. What part was the doctor and what part was us? It’s mostly us—but we need data to prove it.

So that’s the first piece. And to get that data, we’ve tried using chatbots and other engagement tools. So far, we’ve learned that where humans struggle to get other humans to communicate with them, AI chatbots tend to do worse. So we’ve gone back to using simple text message reminders: “Is there anything you need to tell us this week?”—things like that.

As for in-home technology, I’ve gone on a bit of a journey. We tested things like Alexa, sensors, and other tools. That helped us narrow down the data we really need to know if someone’s well or not. I think it’s about identifying when someone is active when they usually wouldn’t be, or when their activity looks different than normal.

I used to think about multiple sensor systems. But recently, I saw a very elegant solution that uses an AI chatbot avatar interface paired with RFID tags in shoes—50 cents apiece. Versus more complex sensor systems, can I get 60% of the data that gives me 80% of the information I need, for 10% of the cost? That’s the goal. There’s just too much data out there.

Fall prevention versus fall detection—that’s key. Anything that motivates people around exercise or engagement, that’s where we want to stay ahead. You can see it in healthcare spending: when someone falls, they start a downward spiral. That’s when all the spending happens. So anything that helps with preventative wellness is huge.

Q: That’s very cool. And you’ve mentioned broadening clinical care, Mike—you were talking about social prescribing. Can you tell us what that means to you?

Mike: Yeah, I just heard that term today, which I think is neat. In Canada, where I’m from, doctors will prescribe National Park passes. In England, they even have a Minister of Loneliness.

The joke I always tell—and maybe one listener will get this—is: “Minister of Loneliness is not the name of the new Morrissey album.” That’s my 1980s joke.

But seriously, I think social prescribing is taking off. I heard a great story on NPR today—Kaiser is supporting an initiative around this. The model is really about understanding your motivations. Why do you want to stay healthy? Why do you want to stay engaged?

What’s really important for people in their 70s, 80s, 90s is to have a strong sense of purpose. Why do you want to stay well? Maybe to see your grandkids grow up. Or to stay in your home. Maybe you want to maintain your garden. The number one reason people move from their home into another home is home maintenance—cooking, cleaning, that sort of thing.

As you age, small frictions become big obstacles. I woke up today and my back hurt—it sucks, but I got through it. For older adults, those frictions increase. Supportive services can reduce the friction, take burdens off their shoulders, and help them return to their baseline.

But you can’t do that without knowing the person and what drives them. Taking your pills every day—it’s not about liking how they taste. It’s about what motivates you.

If we start there and build a partnership, speak to someone in a language they understand—then we can make a difference. I can’t pretend to be a doctor—I don’t know the medical language, no matter how many episodes of ER I’ve watched. But we need to meet people where they are.

All healthcare is local. All social work is local. I think today’s technologies have great promise in expanding these highly successful, local, relational models. That’s what I’m excited to see.

Q: That’s amazing. So as we come to the close of the podcast, Mike—when you look ahead, based on all the innovation and transformation happening in this space—what do you see coming our way? 

Mike: Wow. Well, it’s funny with AI right now, isn’t it? Every 8 months, it seems like there’s a new revolution in capability. I used to be very cynical about it. I’m a marketing professional—my background is in direct marketing.

Then came customer relationship marketing—because consultants needed something new to sell. So I looked at AI and thought, this is just machine learning, right? It’s been around forever.

But when I started thinking about AI as pattern recognition, I began to see the bigger picture. Where else do we find patterns? We find them in nature—in fractals, in repeating structures. AI taps into that same concept of pattern recognition. It’s fundamental.

I don’t think it will ever become sentient—that’s carnival barkery. But I think it has promise. Agent AI is interesting. I haven’t seen it work properly yet—but if we can automate prescription renewals, appointment scheduling, or even coordinating a ride to the doctor—that’s big.

If we can reduce the frictions that prevent people from getting back to baseline—that’s where it’ll be most successful. And most importantly, AI should maximize human time—what we’re doing right now: having a conversation. That’s where the value in healthcare will be—freeing up more time for this.

Q: Absolutely. You know, I think the more help we can get from tools like AI coding, the better. We’re even seeing our clients ask, when we start a project, “Do your people use AI coding tools?” We’re now selecting people who are good at using those tools because some say it can make them up to 100% more productive. 

Mike: And I think there’s a democratization happening with AI. A lot of what I’m seeing is like everyone inventing the same thing at the same time, everywhere. It’s like how the light bulb and the steam engine were invented around the same time in different places—because innovation opened adjacent doors all over. That’s what’s happening now.

But unless you co-create with the people you aim to serve, you have no load around your system. So just a call to action to anyone developing solutions in our space: co-create with the patients. Co-create with the customers. The UX will be simpler, the data will be simpler, and you’ll be far more effective in selling it through.

Q: Mike, on that point, please talk to us about imparting—about your concept of entrepreneur-in-residence. I’ve talked with you about that, and it’s a very successful program. Please share.

Mike: Thank you for bringing that up. I didn’t have it written down, but yeah. In my position, I get a lot of calls from people asking for advice in the aging, longevity, or age-tech space. And by the way, I think we should get rid of the term “age-tech.” I haven’t found a better word yet, but just because you’re of a certain age doesn’t mean you need special tech.

One of the biggest challenges we face is change management. We’re largely a reactionary workforce. We don’t know what we’re walking into every day. If you want to create solutions for our space, you really need to fall in love with our problems before coming up with a solution. That’s part of human-centered design—having deep, embedded experience with those you aim to serve.

So we have a program at our Glenwood community in Marietta, Ohio—a very historic community. We have cottages, independent living, assisted living, and 15 minutes away we have Harmar Place, which is skilled nursing with memory care.

We offer a two-week program where you can come live with us. Week one, you formally shadow different job roles. Week two is kind of a “choose your own adventure.” We give you a persona—like you’re a new resident in independent living. Your persona has specific traits. Walk into the dining room, sit down, feel nervous because no one’s sitting with you.

Our residents are wonderful—they’ll sit with you and talk with you. Make friends. Get to know them. See who they are. If you make enough friends, maybe you can test your prototype or do user group testing. But it’s not going to work unless you can embody the experience and make friends. You’ll learn why people don’t have time—and what conditions make change possible.

Q: Yeah. You immerse yourself in the setting, then pick a problem to solve, and co-create. 

Mike:   Exactly. And we’ve had a lot of fun with it. We’ve been running it for over a year. A lot of the people come from places like New York City. I think half of it is because they like the idea of living in a two-bedroom, two-bath apartment in a peaceful setting for two weeks. But now they’re starting to collaborate and use the experience as a foundation. I love seeing that happen.

I encourage any senior living provider to start a program like this. And any developer—look for these opportunities.

Q: That’s amazing, Mike. Thank you so much. This was a very exciting discussion. Wishing you all the best and hoping to stay in touch.

Mike: I invite everyone to check out unitedchurchhomes.org. For our entrepreneur-in-residence program, the email is [email protected]. We also have our own podcast series—abundantagingpodcast.com—and our Center for Abundant Aging, which champions ending ageism, spiritual wellness (which we didn’t talk about today), and rediscovering purpose. That’s at abundantaging.org.

————————————  

Subscribe to our podcast series at www.thebigunlock.com and write us at [email protected]   

Disclaimer: This Q&A has been derived from the podcast transcript and has been edited for readability and clarity.

Q: Hi, JD. How are you doing? It’s great to have you on the podcast. Awesome. So JD, as you might be aware, this is The Big Unlock podcast, which was started by the founder of Damo Consulting, Paddy Padmanabhan. We’re now in Season 6 and north of 160 episodes. We’ve come a long way since this podcast started.

I’m Rohit Mahajan, Managing Partner and CEO of BigRio and Damo Consulting. Super excited to have you as our guest and looking forward to diving into some topics. Would you like to start with an intro?

JD: Sure thing. I’m JD Whitlock. I’m the Chief Information Officer at Dayton Children’s, a small pediatric health system in southwest Ohio. I’ve had a pretty long career in healthcare IT—30 years now in healthcare. I’m a retired Air Force healthcare administrator.

I’ve also spent time in larger adult private sector systems like Bon Secours Mercy Health, where I focused a lot on data and analytics. Now, as CIO, I do a little bit of everything IT at Dayton Children’s.

Q: That’s great to know, JD. A couple of questions—just curious. What attracted you from being in the military, in the Air Force and Navy, into healthcare, where you’ve stayed for a long time now? And where are you headed? That’s one part. And second, please tell us a little more about your health system. 

JD: Yeah, sure thing. You mentioned Navy and Air Force—yes, I did start out in the Navy. I wasn’t doing healthcare there; I was doing Navy things, driving ships around.

Then I got a master’s in healthcare administration and started healthcare work in the Air Force. The job I had there was mostly in healthcare IT management.

So really, by the end of my Air Force career, I was doing very similar things to what I do today. And a little more on Dayton Children’s—we’re on Epic. We’re big enough to be on Epic and Workday, which I think probably factors into some of the things we’re going to talk about. 

We’re small compared to most health systems. So what does that look like?

It means we have to do a lot of the same things that bigger health systems do, but it can be challenging to have the resources—people and dollars—to get all those things done.

Of course, when you bring your sick or injured child to Dayton Children’s, you have the same expectations for quality and experience of care that you’d have at a larger children’s hospital—like Cincinnati Children’s or Nationwide Children’s in Columbus.

So yes, the challenge is keeping up with larger health systems, but with fewer resources.

Q: I see, I see. And an increasingly difficult environment lies ahead. So I’m sure there are more challenges on the way, and I’m sure the leadership is already thinking about how to navigate those challenges—especially, and we’ll get to that—no podcast is complete without AI. We’ll talk about that in just a moment.

But before that, what I would like to ask you is—you mentioned that you are on Epic and Workday. So please tell us a little bit more about how that drives your innovation, or let’s say, the consumerism from the digital front door perspective. Any initiatives like that?

JD: Sure thing. So in both cases, we spend a lot of money for the care and feeding of those platforms—both in dollars to the vendor and in terms of all the labor that we need to put into them. That’s the bad news.

The good news is we have best-in-class platforms in both cases, and we can do a lot of innovation just by optimizing within these platforms, including some of the generative AI features that both vendors are doing a very nice job implementing into their platforms. That’s very exciting.

We’re early-ish stage with some of that, but the point is—it’s a lot easier to implement these features from within the platform than try to bolt on new things. In some cases, we’ll be bolting on new things, like ambient, and maybe some autonomous coding and some other things.

I’m not saying we won’t do that at all, but probably 90% of what we would do with generative AI would just be from Epic—or I should probably also throw Microsoft in the mix. We’re a Microsoft shop, so we’ll be using some Microsoft tools also.

 Q: Yeah. So could you talk to us, JD, about some of the generative AI use cases that you perhaps are already looking at or might be on the roadmap of these vendor partners that you are going to be adopting? 

JD: Sure. Well, one obvious one is ambient. Most health systems—if not fully in production—are at least piloting or about to pilot something with ambient. I think very soon here, having some ambient solution will be an expectation from providers. And health systems may have difficulty recruiting new providers. And of course, as we have more challenges with physician shortages, that’s going to be a challenge.

One dynamic at Dayton Children’s is, of course, we need to successfully hire pediatric specialists. And they typically are getting out of their pediatric specialty fellowships at large academic medical centers. To convince them why they should move to Dayton, Ohio, we can’t be at a competitive disadvantage to some of the larger facilities. If we are, for example, not using ambient, we’d like to be fast followers. We’re not going to be the first to do things. We’ll leave that to the academic medical centers and some of the truly new things that they’re developing—both on the clinical side and the digital health side.

One of the nice things about being an Epic customer, of course, is there’s such wonderful collaboration between the whole Epic community. If you do something innovative in your Epic build, you go to the Epic conference, you present it, and other people can use that. That’s sometimes what Epic will just build into the next version of Epic. So an awful lot of that goes on all the time. And Epic is rolling out so many new features so fast, it’s actually difficult just to keep up with all the new features that are coming from Epic.

Q: That’s true. It’s a large system, JD. So how do you separate the wheat from the chaff? That’s something we were kind of hitting on before we started the podcast. What are your thoughts on that? How do you decide what is critical and core, and what can be done later or perhaps doesn’t need attention right now? 

JD: Sure. So as a general concept—just good governance, right? And not chasing after, as we like to call them, the “bright, shiny objects.” Even with core generative AI, you’ve always had that problem. Somebody goes to a conference, they see something that looks cool—and it may be cool—but there’s not enough return on investment to spend the dollars we don’t have on that thing.

So we’ve had that challenge for a long time. I would say generative AI has ramped that problem up a few notches because there’s so much hype. You have to be careful—not just about wasting money, but also the additional considerations that come with generative AI that we didn’t always have with other things. Things like ethical considerations and medical-legal concerns. So we need to pay a lot of attention to that.

I try to stay up on all this, of course. And when I listen to very smart people who spend their entire lives focused on generative AI, they often talk about the investment bubble. Two things can be true at the same time: One, there’s amazing science and capabilities advancing very quickly. And two, a lot of the investment money pouring into this is going to be bad investments because nobody’s going to pay a gajillion dollars for that thing you built.

So that’s where you have to be very careful. Now, how do we handle that? Well, we handle it like we always have—by asking hard questions about ROI. And sometimes we do things that have more soft ROI than hard ROI.

Ambient is a great example. Reasonable people can disagree about the hard ROI, but there’s really no question about the soft ROI—keeping our providers happy. You hear story after story: “I was about to retire early,” “I was burned out,” and “this really brought back the joy of practicing medicine.” Pretty much every system that’s implementing ambient gets dramatic stories like this from providers.

Soft ROI is important too. You just can’t buy everything that has soft ROI—you have to be judicious.

Q: We had touched upon using some of the new tools that are coming out for enabling coding. What are your thoughts on some of these tools, JD?

JD: Yes. This is an interesting space. It may be something we work on with additional vendors. In fact, we’re about to go live next week with Epic’s professional billing—what I believe is called the DB Coding Assistance. It’s a lighter-weight AI solution aimed at making our PB billers’ and coders’ lives a little easier with some tools from Epic.

There’s a spectrum of billing complexity—from professional billing to hospital outpatient and inpatient. From what I understand, inpatient is still too complex for full autonomous coding. But in the hospital outpatient space—that middle ground—autonomous coding, thoughtfully applied, can really help our coders and billers be more efficient. We’re exploring some of those vendors to see if there’s a good fit for us.

Q: That’s great. So, as a smaller health system, how do you approach innovation? How do you keep up with the larger systems and still deliver quality care? 

JD: Sure. Something else—there’s a term commonly used in the Epic ecosystem: “imitate to innovate,” right? If you can get past the concept of not being proud about implementing something that somebody else developed someplace else—that’s really the answer. We like to say we want to be fast followers. Most people in IT are familiar with Gartner’s hype cycle—the peak of inflated expectations, the trough of disillusionment, and the plateau of productivity. We’ll let others go through the trough of disillusionment. We want to be there for the things that actually work.

We’re not just rolling the dice on whether something will work. No, that worked. This thing worked at another children’s hospital. And we know those people—we have really good relationships with pretty much all the CIOs and CMIOs at the other children’s hospitals. We go to conferences and talk to each other—“Oh, that new generative AI feature from Epic worked wonderfully for us,” or “that one didn’t work so well—it wasn’t a good match for pediatrics,” or whatever the case may be. We talk to each other and increase our confidence. Nothing’s ever 100%, but we’re more confident that it’s worth the effort.

Q: Right. And JD, you’ve been at health systems that weren’t pediatric-focused as well, right? So what’s the difference? I’m curious—in the world of pediatric hospitals, how are things different compared to other health systems?

JD: Sure, thanks. Some things are different with pediatrics. It’s unfortunate, but it also just makes sense—the way the world works. When you have innovators and venture capital funding innovation, a lot of the dollars go to adult medicine because that’s where more of the money is. Pediatrics sometimes plays second fiddle.

Maybe Epic rolls out a new predictive algorithm that works better for adults than for pediatrics. That was true for the sepsis predictor. I remember a pediatric CMIO talking to me about why that was. So we just have to be cautious.

Other examples—imaging. Now we’re talking more predictive than generative AI. Some of these are technically generative, but there’s a lot of FDA-approved, highly effective new imaging tech powered by AI. I was talking to our radiologist about that, and at their conferences, they’ve noticed there hasn’t been much for pediatrics yet. A couple things—bone age prediction, maybe one other—but that’s about it.

So sometimes we just have to wait. In other cases, there are people doing innovative things targeted at pediatrics. We’ve been looking at a couple of NICU-focused solutions—for a better parent experience. Your precious little new baby, sometimes very tiny, is in the NICU. You have to learn a lot quickly—talk to the doctors and nurses and figure out what that all looks like. How can we make that experience better?

Also, we’re an ACO—we want to make sure we’re spending our dollars wisely. We want kids in the hospital when they need to be, and home when they can be. Some solutions around tube feeding, oxygen—where we can send infants home earlier than we otherwise could with better remote monitoring and communication tools. In some cases, there’s real innovation going on that’s very specific to pediatrics.

Q: That’s great to know. You also mentioned Workday, along with Epic, as a major system. Have you seen anything on Workday’s roadmap that you’re considering? 

JD: Yes, we were just looking at this yesterday in our Workday executive governance meeting. We asked our account team to put together a chart of all the generative AI features Workday has. There were a lot—it all had to fit on one slide with pretty small fonts. We color-coded them—what we’re licensed for and using, what we’re licensed for but not using yet, and what we could be doing but aren’t licensed for.

One big difference between Epic and Workday when it comes to AI: Epic has a deep partnership with Microsoft. That’s where the generative AI and cloud compute happens—in Azure. Epic is hosted on-prem for us. A lot of Epic customers are still on-prem. Workday, by contrast, is built with modern cloud architecture from the ground up. They don’t need to partner with anyone—it’s just built into the platform. Both vendors are doing AI differently based on their system architecture.

Q: So again, JD—just a curious question because I’m trying to build a picture in my mind. Let’s say Epic and Workday are two major systems that are clearly top of mind. Are there two or three other systems, not in the same space, but in different domains, that are also driving your AI or GenAI roadmap?

JD: Sure. For health systems, another very strategic area is your PACS vendor. For modern PACS vendors, you want them to plug into all the AI tools that are coming. In some cases, they can do that natively. In others, there’s middleware that adds AI features. You want to be able to add that capability easily.

Then there’s the vendor-neutral archive—getting all the images in one place. We haven’t always done a great job with that. If a radiologist ordered and read the image, it went into PACS, but other images—ordered and read by other providers—sometimes get squirreled away. That’s not ideal, especially if you want to apply AI across images. A vendor-neutral archive is typically a better architectural solution.

One of the most strategic acquisitions we’ve made in the last seven years was PACS. When I got here, we were already on Epic, and Workday was just starting. But PACS—we were replacing an old legacy system. I knew we needed a PACS and VNA combo that would be future-proof. We selected Sectra, which has long been considered best in class, and we’ve been happy with them.

There aren’t a lot of pediatric-specific AI tools yet, but we have good confidence in our direction. We’re doing well getting the other images—like point-of-care ultrasound—into the VNA. So as AI tools become available, we’ll be able to implement them easily.

Q: You mentioned the vendor-neutral archive, and that reminded me: for AI, we need data, right? And we need data engineering on top of that. But data’s often siloed—Epic has its data, Workday has its own, PACS has its own. How do you approach enterprise-wide AI that cuts across systems or functions? 

JD: Good question. If we had more money and more time—and maybe if we were a big academic medical center—we’d be spending more on a true enterprise data warehouse. Honestly, we don’t really need that at Dayton Children’s. The vast majority of our reporting and analytics lives in Epic.

That said, we are dabbling with Microsoft Fabric, because that’s where Epic is headed for the cloud. We’ve been doing Power BI for a while. For our enterprise scorecards, we bring in Epic data, Workday data, and other sources to build dashboards for leadership.

Another example—patient survey data. We take survey data from our vendor, mash it together with Epic data, and build dashboards. That helps us drill down on metrics like Net Promoter Score, which we’re proud of. And we’ve done analytics around that—asking why certain patients in the ED at certain times aren’t satisfied. You can’t tell just by looking at the survey data—you need the Epic data too. Then you can better implement fixes.

Q: That’s very thoughtful and insightful. JD, I know you mentioned that you also do some consulting. Can you tell us more about the kinds of engagements you take on and what excites you? 

JD: Yeah, I do a little consulting on the side, after hours. I joke with my boss that it makes me a better CIO—it helps me stay connected. I talk to people—sometimes investment companies—who are trying to decide whether to invest in a particular software solution. They want to talk to someone actually using it.

Sometimes people have a new product idea and want feedback from someone like me. I see things coming into the market—vendors acquiring tools and building platforms. Most of the work I do is one-off expertise. I’ve been doing this for a long time.

Occasionally, I help tech companies on an ongoing basis—helping with go-to-market strategy and how to sell into health systems. That kind of thing.

Q: Thank you, JD. As we wrap up, are there any other thoughts or things you see coming in the future that you’d like to share? 

JD: One last thought—it always comes down to good governance. I talk to other CIOs and CMIOs who struggle with too many requests. “How do we deal with all these requests?”

Something that’s important—but not fun to talk about—is acquisition policies and governance. People are getting wrapped up in all the new GenAI stuff. You need to think hard about the ethical and medical considerations and build those into your evaluation and procurement processes.

What I see a lot of people doing is starting 17 new AI committees. And I think, who has time for that? It’s better to work AI into your existing governance structures and procurement policies. But those policies have to have teeth. You can’t let everyone buy any IT thing and toss it over to IT to implement.

Then IT ends up unable to do the core things we’re supposed to be doing because we’re trying to plug in new tools that don’t fit the architecture.

Q: Understand. Thank you, JD. This has been a great conversation—really appreciate it.

JD: Thanks for having me. Hope we can continue the conversation in the future.

————————————  

Subscribe to our podcast series at www.thebigunlock.com and write us at [email protected]  

Disclaimer: This Q&A has been derived from the podcast transcript and has been edited for readability and clarity.

Q: Hi, Alicia, welcome to the Big Unlock podcast. It’s great to have you here. This podcast is now in season six, Alicia, and you’re looking towards an exciting conversation here. I am Rohit Mahajan. I’m the Managing Partner and CEO at BigRio and Damo Consulting, and would love an introduction from your side. 

Alicia: Hi, Rohit. Uh, lovely to be here, and thank you so much for the invitation. Well, thank you. Yes, I’m Alicia Abella, and I am the AI Product Lead with Novo Nordisk. I joined Novo Nordisk in November of 2024. I will say that I am new to pharma and was very intrigued by the opportunities that pharma has for AI. My whole background since graduate school has been in AI in some form or fashion, so we could talk a little bit about that and where I am. And I’m from Morristown, New Jersey.

Q: Absolutely. That is awesome that you’re new to pharma and you’re kind of looking at it from perhaps multiple different experiences that you already have. So please tell us, Alicia, how did you get started? What drew you to AI and now to pharma, and what are some of the experiences that you’ve had before you got here? 

Alicia: Yes, happy to do that. I would say that my involvement with AI started rather serendipitously in graduate school. I was a PhD student at Columbia University and I was looking for an advisor—specifically a tenured advisor—because I knew as a student that if you signed up with a tenured advisor, you wouldn’t run the risk of them not getting tenure and then having to move to another university to follow your professor.

So the professor I worked with on my thesis was in the area of computer vision and image processing. I also had a co-thesis advisor who was in natural language processing, so I had a very early experience in that area that now is so popular with large language models and AI. This was a period where the technology was using different techniques, algorithms, and approaches than we do today. But still, there was that sentiment that artificial intelligence could be used to do many things and assist people in discovery.

The actual work I did for my PhD thesis was in the life sciences and healthcare industry. Ironically, even though I’m new to pharma now, I feel like I’ve come full circle. My thesis involved developing a software system that could do image processing on radiographs of human kidneys and the urinary tract system—automatically detecting calcific densities and kidney stones—and then automatically generating radiology reports.

Part of my thesis was to see how well a machine could do that compared to a radiologist, and it turned out it did pretty well. And this was in the mid-nineties.

That experience propelled me into AI, and especially the work I did on my thesis in natural language processing led to an opportunity at AT&T Bell Labs. I got a job in their research organization specifically doing speech and natural language processing research. What I think most people don’t realize is just how long research in this field has been going on. We often think that large language models and ChatGPT just burst onto the scene, but in fact, the foundational work has been going on for decades.

When I joined Bell Labs in the mid-nineties, they had already been doing speech and signal processing research for nearly half a century—40 to 50 years. That’s a humbling experience because you realize just how difficult these problems are.

So that was a challenge, and I spent a lot of my career—25 years—at Bell Labs doing many different things. But the part that still resonates with me is the work I did initially on spoken dialog systems for customer care. 

So that’s a little bit about my initial beginnings with AI. And then yes, after a long career at AT&T, I took on a role working at Google. Because of my long career at AT&T, Google—at the time, around COVID in 2020—was standing up an entire organization devoted to different industry verticals. They were hiring a market lead, a managing director for the telecom, media, and entertainment industry vertical. I was recruited for that role and joined Google in August of 2020, right in the middle of the pandemic.

It was interesting because Google Cloud at that time was growing very rapidly. They really wanted to go after some of our strategic customers in the telecom space. Having spent so much time at AT&T, I could talk the talk. I’d walked in the shoes of our customers. My role was really to talk to our most strategic senior executives and C-suite executives across all of the Americas in that industry to try to understand: What are their pain points? What problems are they trying to solve? And how could Google and Google Cloud’s products and services help them solve those problems?

In my last year at Google, before joining Novo Nordisk, I was the Global Practice Director for AI and Machine Learning. I was essentially the bridge between our go-to-market organization and our product and engineering teams that were developing the AI products that are now very much in use—Gemini, Vertex AI, and others. I was ensuring those products and features were really meeting our customers’ needs globally and across all industries.

So I went from telecom to representing all industries and from the Americas to a global role. I spent about a year in that role until I got a call from Novo Nordisk. As I mentioned earlier, I was very intrigued by the opportunity to take AI and apply it to an industry that I think has tremendous potential—to apply AI across all of its business functions.

We often hear about AI being used for drug discovery, and pharma companies are very much involved in using AI to help accelerate drug discovery. But there’s also an opportunity across commercialization functions as well, which is where my current focus is at Novo Nordisk—to bring AI to the commercialization space.

We’re trying to move from traditional commercialization techniques to thinking about how we can use AI to accelerate the work that needs to get done—whether it’s marketing campaigns, legal issues, HR—you name it. The entire organization involved in that go-to-market aspect of taking a drug, once it’s been discovered and approved, to market.

How can we use AI to accelerate that process and make it better? Make it more personalized? How do we find the right patients? There are so many applications in the commercialization space, and I think we’re only scratching the surface.

Q: Absolutely. Very exciting, Alicia. Thank you for sharing that.
As you go about this role, we talked earlier about AI adoption and change management—how do you get people to embrace it? What are you seeing in the business enterprise, and what are some of the things you’re doing to make this happen?

Alicia: That’s a great question, Rohit. When I first joined Novo, I spent the first few months—maybe a good solid three months—just going around and talking to various leaders across different business functions to understand what they were doing, what their current sentiment was, and what their understanding of AI looked like. I needed to understand where Novo Nordisk was in that journey.

It was varied. There were folks who were very excited about the prospect of AI, and others who were afraid of it—or still are—partly due to a lack of understanding and education, and partly because of the compliance and regulatory risks they know or have heard about. We’ll probably get into that later in the podcast.

I also created a survey at the time, which I sent to the marketing teams to assess their general understanding of AI and create a baseline for myself. Through that process—conversations and survey results—I realized there was a need to demystify AI for many people. I also needed to develop very strong relationships with our legal and compliance departments, to involve them very early in any AI solutions we were thinking of developing. That way, everyone would feel safe and confident that what we were building wouldn’t create risk for the company.

So I came up with and launched, just two weeks ago, an AI Ambassador Program. I wanted to find and engage the people who were excited about AI, wanted to learn more, and wanted to be part of a community that could share that knowledge with their peers.

It was important to me that these ambassadors represented all job functions within the enterprise because they know their day-to-day work better than I do. They know where AI could be applied. They could become a kind of flywheel for me within their own organizations.

I put out a request to the organization for volunteers, and the response was overwhelming. I surpassed my expectations in terms of the number of ambassadors I hoped to recruit, and now I have a big cohort.

Now the real work begins—how do we equip the ambassadors with the knowledge and education they need? My goal is to awaken their curiosity about AI and inform them in a way that is relevant to their context. I want to give them a broad understanding of AI, what’s out there, what’s coming, and what’s on the horizon. Get them excited. Get them thinking about how to apply AI to their day-to-day work so they can bring that knowledge back to their peers.

I think it’s really important that this happens peer-to-peer. It’s not coming from a top-down directive that says, “You must do this training.” That kind of approach doesn’t work—especially if it’s not connected to their day-to-day work. That contextual understanding is critical, and that’s what the ambassadors can bring. I can help supplement it by bringing in the outside-in perspective of what’s going on in the AI landscape today.

We also have internal tools—Novo Nordisk ChatGPT, for instance—that employees can use to query and ask questions. It’s all within compliance and legal, so it’s a safe place for them to experiment. We also have Microsoft Co-Pilot, and there’s a lot of training available there too. But my focus is to expand their thinking with a broader understanding of AI.

We launched the AI Ambassador Program last week. We meet monthly to discuss different AI topics, and I know the next topic we’ll cover will likely be use cases that are relevant to the organization. We already have a lot of AI use cases across the company—including globally—so knowledge sharing is a big part of this program too.

Q: That’s wonderful. It’s a very unique approach to increase adoption, Alicia. You just mentioned use cases—what are some of the ones you’re already seeing, or that you’ve seen other pharma companies pursue? 

Alicia: Sure. Since my focus is on commercialization, I’ll highlight use cases in that area. One of them is what I’d call using AI for knowledge search. One of the first things I noticed was how much data our marketing and insights teams have access to—whether it’s reports they generate themselves or third-party vendor research to understand how our products are doing in the market and what our competitors are doing.

There are so many disparate data sources and documents. It’s hard to find what you need. So, we’re working on using generative AI to provide a conversational interface where market researchers can ask questions, and the solution can sift through thousands of pages to return useful responses—with attribution, so they can validate the answers by going back to the original documents.

That’s one use case—market research and competitive intelligence using knowledge search.

Another is content generation. We’re using generative AI to come up with variations on ad messaging and new campaign ideas. We’re also exploring image generation and short video clips to help marketers communicate their ideas to ad agencies faster. Using GenAI in this way can help accelerate time-to-market for campaigns.

Ultimately, the hope is that we could use these technologies ourselves to create content, instead of outsourcing it.

Then, more traditional AI techniques are being used to analyze large datasets to understand healthcare providers—their habits, the types of patients they’re seeing, and who they’re diagnosing with conditions our products support. These insights help us better position our messaging and outreach to HCPs.

It’s a powerful use case that helps us reach more healthcare providers, who can then reach more patients. That’s how we ultimately expand access to our therapies.

Q: Great use cases, Alicia. Early on when we were talking about a product mindset and you know, a very disciplined way of approaching the implementations or the use cases itself. So, could you tell us a little bit about what does that mean?

Alicia: Yeah. So, when I was first recruited to join Novo Nordisk, I thought the part that intrigued me, in addition to it being a new industry for me and a great learning opportunity, was this idea of applying the product mindset to developing AI solutions for the pharma commercialization team. Because I’ve seen it too often in my history of being in technology and development, and being around technologists, that they get very excited about an idea—an idea for a product—and they go off and build it without actually taking into consideration: should we be building it?

Just because you can build it doesn’t mean you should build it, right? There’s a lot of work that goes on in evaluating and determining whether an idea should actually be productized.

So my role was to bring my experience—all the way back from my AT&T days and Google days—around a product management lifecycle mindset. Think about product management as: design, develop, test, monitor, iterate. Design, develop, test, monitor, iterate—how to do that, and how to put that kind of rigor into decision-making about what AI products and solutions we should be developing.

That’s what I’ve brought to the exercise of us picking what AI products to focus on. Because there’s limited resources, so we have to figure out and prioritize.

Step number one is: let’s understand the user need. Let’s understand and talk to those end users—those key stakeholders and sponsors—and understand the problem they’re trying to solve. Figure out if indeed you need AI to solve it. Because in some cases, you may not. So that’s part of the process.

If indeed AI is a good tool for it, then what’s the business value? Can we create a business case for it? What are the implications of building this solution in terms of compliance, risk assessment, technical feasibility? All the things you have to consider to generate and create a new product.

Bringing that process here helps us focus on developing AI products that we know will create the biggest impact and have the most value across the largest set of stakeholders.

And so that’s kind of what I’ve been bringing to Novo Nordisk and to our AI product experience—because I think that’s maybe something that any company needs. And yes, pharma is still relatively new to AI, especially in commercialization. And I think a little bit of governance, without it being too heavy-handed, can help drive the innovation and drive it in the right way for the right problems.

Q: Absolutely. And in pharma, it’s highly regulated, as we all know, Alicia, and there’s a lot of compliance. How do you tackle that aspect of it in the journey? Could you tell us how you’re approaching that as well? 

Alicia: I’m glad you asked that question because it’s one of the things I mentioned earlier about my listening tour when I first joined Novo. When I was talking to different folks, I made sure that legal and compliance were on that tour. I wanted to understand Novo Nordisk’s guiding principles and how they were thinking about AI.

We have AI principles and guidelines that we follow today. And what I’m currently involved in is working very closely with our data ethics, compliance, and legal teams—very early on.

I told them, “You’re going to be my BFF,” right? You’re going to be my best friends. Because it’s important to make sure that what we’re developing is done in a compliant manner.

What I can bring to that process and that team is an understanding of where to put those guardrails so we don’t stifle innovation. We still need them, but we do it in such a way that we manage risk while still being able to innovate.

Having a strong relationship with that team—where we both understand what we’re each trying to achieve—is important. I bring them in very early, even when we’re still just thinking about an idea for an AI solution. I say, “Look, this is what we’re thinking—are there any big red flags I should be considering?” So we can address those right at the beginning, before too much effort and resources have been devoted to developing a solution that they might later come in and say, “Oh, you can’t do that.” We definitely don’t want to do that.

That’s an important component, especially in a highly regulated industry. And I think there’s still a lot of room for innovation, even within those boundaries.

Q: Great. So, as we come towards the end of the podcast, Alicia, I’d like to ask—what are your thoughts or ideas about future trends? What do you see from your perspective? 

Alicia: I wish I had a crystal ball! But if I look into it, one area that I think will be very interesting for AI in the future is making sure that we’re marrying the AI fundamental technologies with the user experience.

Part of that is driven by my entire career—I’ve always been focused on that user-centric view. To drive adoption of any product, you have to make it usable. You have to make the experience something that people will want to use—something intuitive, easy to use—that will drive adoption.

It obviously has to solve a problem. Assuming it is solving a problem, it should do it in a way that makes it easy to interface with. I think part of what made ChatGPT so prolific in terms of adoption was how simple that interface was. It’s just a window that says, “Ask me a question.” You just type it in the way humans are used to asking questions.

So I think as we build wrappers and layers on top of these fundamental large language models, it’ll be important to ensure that simplicity of user experience remains.

That will be a trend going forward. I think one of the big tech giants creating large language models will now focus heavily on user experience.

Maybe I’m just channelling my experience when the iPhone first came out. Steve Jobs had that focus and fascination with experience. It was all about the experience—being very user-centric. I think we can’t lose sight of that.

Another trend I think we’ll see is large language models evolving beyond just text, image, and video—to start bringing in more contextual knowledge. The kind that humans bring. That’s the missing link right now.

It’ll be interesting to see what the future holds on that front. I’m a big Star Trek fan—especially The Next Generation. There’s an important character named Data. He’s a machine, an AI. All he wants to do is be human.

All of his attempts at being human—wanting to be an artist, a writer, a pet owner—everything is about the machine wanting to be more human.

I think we still have that desire—to see how we can make these machines behave more like humans, without taking us out of the loop. Of course, there’s that fear too, but that’s a whole other podcast episode, Rohit.

But there’s a lot to be excited about in the future, and I feel fortunate and privileged to be part of this experience and to see where it all unfolds.

Subscribe to our podcast series at www.thebigunlock.com and write us at [email protected]  

Disclaimer: This Q&A has been derived from the podcast transcript and has been edited for readability and clarity.