With all the buzzwords surrounding artificial intelligence, from smiling promotions to angry chagrins—promises that it will revolutionize how we work and play, while “slop” becomes Merriam-Webster’s word of the year—it can be difficult to get a clear view on what AI can bring to patient care.
When it comes to this regulated industry, where life and death risks come into play, hype quickly gives way to a more sober assessment of the landscape, and there’s less of a rush to throw AI at every possible use case to see what sticks.
Still, the FDA has issued more than 1,350 green lights to AI-powered tools, and its pace is accelerating. More than 250 were authorized in the first nine months of this year alone, compared to 235 in all of 2024.
The lion’s share, of course, has been in radiology: certain AI approaches lend themselves particularly well to analyzing X-rays, CT scans and other images, and seeing the patterns associated with disease.
But Bilal Muhsin, president of BD’s connected care division, sees AI as being capable of much more. The company is chasing a virtual model of a patient’s physiology, which will allow it to track which medications go in and what vital sign changes are expected to show—helping clinicians to respond faster and smarter.
Muhsin took the company’s newly created role this summer, after 25 years at the patient monitoring giant Masimo. His unit was restructured to include not only BD’s AI efforts—which recently launched a new platform—but also its robotic systems for filling prescriptions at hospital pharmacies and other networked hardware.
We had the chance to talk about how BD wants to become more of a service provider, by using AI to link every one of its devices, and put an electric eye on everything from pulse rates to infusions and pills.
Conor Hale: First, I’d like to ask you about where you see AI going in healthcare. Is AI just old news by now? What’s left to talk about, at this point?
Bilal Muhsin: Good question! So it’s been called different things, right? We started calling it algorithms, and then that got boring—so we went to AI, and we added a lot more data to it, and allowed the machine to do a bit more with the computing power. And what people are talking about, when you look at AI overall, everybody's talking about where data centers will be and how it will help us do everything in the future.
But I don't see it as old news. In healthcare, there's a lot more to do with AI.
I think it's at a place where people are comfortable talking about what it can do for imaging, but the human body is complex, and I think once you start to look at the physiological implications of what AI can do, I think it's a little bit more of a difficult question.
I think that's what differentiates what we're doing, on the medical device side of things, versus what Silicon Valley is doing with AI.
Silicon Valley people are just throwing a ton of data at it and expecting it to solve. But I think the reality is that there needs to be AI within certain bounds when it comes to healthcare. I think you need to build that physiological model first, and then allow AI to play around [with] it. So I think there's still a lot more to do.
CH: Yes, absolutely. And they're not just throwing a lot of data at the AI, they're throwing a lot of AI at different use cases, and trying to find something that sticks.
BM: Sure.
CH: With that in mind, and talking about the regulated medical device space, what are you most excited about seeing in the coming year? And I know you joined BD about six months ago, so what's on your new whiteboard for 2026?
BM: So first, AI is not new to us. Talking about the connected care segment that I run, which is basically all of our digital platforms, that includes our pharmacy automation systems.
I view most of what's about to happen in the healthcare world as probably starting to change financially... There's going to be a little bit more of, ‘I believe you, but you got to prove it.’
Think of robots that do the actual packing of pills, and how it all gets put together in pharmacies, and then sent to the hospital floors where dispensing and infusions happen, and then patient monitoring—all that encompasses what we call connected care.
And when we look at that and say, okay, what's coming right now? Well, we've been doing AI quite a bit. Our advanced monitoring platforms have some of the first AI-enabled tools that are predictive in nature for the physiology of patients. We can predict when they're going to get hypotensive. We can predict whether autoregulation is intact or not. All of these are pretty advanced tools that we're providing to clinicians to provide better care.
We're taking that to the next level. So instead of these individual products or platforms, now we're adding an AI layer above them all. We've just launched what we call BD Incada, our cloud-based AI platform that's initially deployed on one of our very famous drug-dispensing platforms, which is the Pyxis Pro.
We're looking at the data and seeing what natural language processing can do. So instead of generating a bazillion reports, where you're customizing and so on, you can type in plain English what you want from it.
So that's just the beginning of it, but what BD Incada is going to do—in this coming year and years to come—is we're going to start to push each one of our platforms underneath it. So we’ll be able to look end-to-end, in terms of medication management and what's happening in hospitals.
We'll be able to say, hey, you don't need a clinician clicking every two to three minutes to react while monitoring a patient while you're trying to infuse something. These things can talk to each other, and the value can be much greater in terms of patient outcomes, in terms of clinician timing, and more. So that's really what's exciting with the launch of BD Incada.
CH: So do you see this more as a tool that clinicians will be interacting with on a daily basis, or will it all be operating more under the hood?
BM: So I think when it comes to connected care, a lot of companies out there are talking about what they can do with data—and they're saying, hey, we can give you visualization tools, we can notify you, we can create some smarts around it.
What we're here to do is really try to achieve outcomes, whether it's clinical outcomes like reducing lengths of stay, or reducing bounce backs and so on.
But what we're going to be doing, initially, is we're going to have the clinician in between. We're not going to go to a closed loop right away. That takes a while.
So what we're going to do is create applications where clinicians can optimize their time. So we'll make a recommendation on what we see in the patient’s physiology, versus what you're trying to achieve in an infusion, for example. We can say, ‘We can automate a protocol for you. Does this look good?’ If yes, we will then take on the task of doing that.
In the future, it may be a completely closed loop. We'll ask the clinician, ‘What are you trying to achieve?’ Great, we'll look at the monitor and the titration and automatically do all of it here.
But initially, we're going to keep the clinician in between and keep them educated, because we want to take them through a journey.
Because there's a bit of a trust factor, right? If I came to you five or seven years ago, and said there's going to be a driverless car that's going to come by and pick you up, and it's going to deliver you somewhere, you could say, ‘I'm not getting into that car.’
But today, people are more likely to get into that car, and maybe because they're already driving a car that has autonomous driving capability. They’ve used it intermittently, they start to trust it, and they build up to it, and that's what we're going to do in healthcare over time.
CH: Do you believe that the recent advances in AI were necessary to build a platform like Incada? The healthcare industry has labored for years on issues such as interconnectivity and data standards. When you’re talking about different types of devices talking to each other, does something like generative AI help smooth over those bumps?
BM: Let me put it to you this way. One of the reasons I joined BD is it's probably the only company in the world that can achieve what we're talking about, because it has the breadth across the platforms that we're discussing.
When I'm talking about what's happening in pharmacy automation, or what's happening in dispensing, infusion and monitoring—that's all still within the ecosystem that BD controls, which is a big advantage for us. And in each one of these product categories, we have a significant and probably leading market share.
I've been in this space long enough to see companies say they're looking at ways of sharing data and interoperability—and I've attended HIMSS for the last 10 years, and everything like that—and it's gotten nowhere.
And the reason why it's gone nowhere, to be honest with you, and all companies have a portion of this, it’s because people didn't know what value the data was going to bring. And everybody was concerned about whether they’re giving up the value by just adding interoperability.
The way to think about it has to be different, to tell you the truth. The way to think about it is, it doesn't matter whether you're passing up your data—‘who's doing what’ with the data is what’s going to bring value to the end user at the end of the day. And tools are improving, and AI is improving, and I think people can't wait anymore.
I think the expectation today, when you talk to clinicians, is they don't care if you're bringing things to dashboards or doing just very little interoperability by passing data to the electronic medical record—that's something that's a given now, right?
They can see what's happening in other industries, and the expectation today in healthcare is that it should move as fast. People interact with their phones, interact with ChatGPT, or any of the AI models, and they know the capability. So I think the pressure on the industry is now to start to deliver on the true outcomes that AI can bring, and not sit here and still wonder, hey, should I share my data or not share my data?
CH: Building on that—what are your thoughts about the disconnect, or maybe the distance, between how AI and machine learning's promises have been perceived in the healthcare industry, going back years, versus the more recent developments in large language models? Like you said, there's a trust factor involved.
BM: First, in data accessibility—look, let's be honest, data going to an electronic medical record is a snapshot of a certain segment of data every 15 minutes. That doesn't bring much value when you're looking at a physiological model of a patient and looking for changes in health.
So let's take that first step, before the trust factor comes in. And what we're talking about with BD Incada, is we're going to be a near-real-time system. Once you get the quality and the frequency of the data right, then you’ll be capable of having predictive models that will have clinical impact.
Now, the trust factor, in terms of what is possible, our approach will be a ‘crawl, walk, run’ approach. If we start with just notifying clinicians of things to watch out for, beyond what's already being done, that means more alarms, right?
You’ve got to consider alarm thresholds, because that's typically what happens in a hospital. You hear something beeping, and red lights are flashing—that, to me, is reactive. There's a physiological change that just happened to a patient, and now the machine is telling me it happened, and a clinician is going to react to it.
But if we—with these datasets, and with our AI and cloud capabilities—start to tell clinicians that there's a likelihood of a deterioration, or an event that will occur in the future, when they see this flag come up, clinicians will start to say, okay let’s watch. And if they wait and they see these events start to occur, then they start to trust more of what the data is telling them.
So we'll start there, by showing them what the AI is seeing, and then showing them why. I think that's very important—just telling somebody blindly that something's going to happen, that's one thing. But if you tell them why, that you're seeing instability in x, y and z, then clinicians start to understand how the AI model looks at physiology.
I think, after they will start to build trust with these platforms, we can get to offering a guided protocol… and then we get to semi-closed loop, or closed loop, where it's operating with the clinician still being able to override, but taking it into a bit of an autopilot modality.
CH: Can you tie that more to what you said about automating infusion protocols, and how Incada’s first steps with the Pyxis Pro will be with oral medications such as capsules and tablets? Were those systems wholly siloed before, and where do you see new opportunities?
BM: Yes, so absolutely they were siloed. Even though our previous Pyxis had a system for hospital inventory management, where you can see whether there was diversion and so on—it wasn't connecting the medication that's being dispensed to the patient. It would lose that context.
Now, once you have the monitoring for that patient also connected to Incada, we can say, there's a dispensing that's occurred, and there's these pills that are about to go in. For example, if an oral opioid is about to be provided to a patient, we can immediately trigger an opioid monitoring algorithm. Or if there is a certain cancer medication that's about to be provided to a patient, we know there's going to be a physiological change, and now the monitor can anticipate that react differently or warn differently.
I've been in this space long enough to see companies say they're looking at ways of sharing data and interoperability—and I've attended HIMSS for the last 10 years, and everything like that—and it's gotten nowhere.
So now whether the dispensing occurs from the Pyxis Pro or from an infusion, we can immediately look at that physiological patient model and assume what changes are about to occur—because that's what clinicians try to do in real time, right?
So we're gonna be the only company that will be able to see what's going into the patient, whether it's from the dispensing or infusion, and then what's happening to their physiology in near real time, which is really exciting.
CH: So that’s definitely a step above just monitoring a patient.
BM: Yes, it's huge. Any time you're trying to do a closed-loop system—let's take diabetes, for example. Imagine if I came to you and said, we'd like you to build a closed-loop diabetes system, by only having data from the [continuous glucose monitor] and not the insulin pump—you’d say I can't do that, I need to know what I'm putting in and how the body is trying to react.
So if you take that model—a CGM and a pump—that's literally what we have here. But moreover, they're also trying to see what the person eats, for the full picture. You can think of all that in terms of drug dispensing: what oral medications did we give them, what intravenous medications did we give them, or what fluids we're providing them, and then how should that patient react? How should the different organs react? And is all that within the normal realm, or are we seeing issues?
So if I'm trying to manage somebody's blood pressure, and it’s high, I'm going to give them medication to reduce the blood pressure. But maybe we don't see the blood pressure start to drop, or we see a reverse effect. We can immediately stop everything and say, hey, you need to look at that. Or—if it is going in the right direction—we don't overreact to blood pressure changes, because we know there's a drug that's trying to put it within a certain range.
That will be the smarts in the system. At the next level, we’ll be able to look at everything that’s being inputted and what's being outputted from the physiology of the patient.
CH: Shifting gears back to what we touched on a little bit, when it comes to Silicon Valley. There are a lot of worries about AI-generated slop, or messy hallucinations. Are you worried at all about the impacts on healthcare and the life sciences industry—or are you worried about the promise of AI in that space being tarnished in some way? Do you see it as a reputation management issue, or are there other things that companies can do?
BM: I think there's a bit of a difference because the audience is so different. I think when you're targeting the consumer world, a lot of people may be fooled and taken for a ride a bit.
Here, the audience is really smart and really savvy around costs.
CH: And skeptical.
BM: Yes, and skeptical. So they're not going to be willing to pay for something that doesn't achieve the outcomes they're targeting.
Can BD do a solution sell versus a product sell? That's the first step to capturing that SaaS model. Look, this will be a journey.
Actually, I view most of what's about to happen in the healthcare world as probably starting to change financially—meaning the whole idea of me just selling capital and a promise is probably going to reduce quite a bit in hospitals.
They’re going to move into a [software-as-a-service] model, and they're going to want a lot more risk sharing-based approaches—in terms of, if you say you can do this, okay, then here's the cost savings on my system, and if you can achieve this, then I will provide you with the benefit as well.
So there's going to be a little bit more of, ‘I believe you, but you got to prove it,’ and you’ve got to show the value before they pay enough money for it.
I think what will happen is that only the companies that are able to deliver on those outcomes will be the ones that succeed. And if they're producing—as you referenced, garbage in, garbage out—even if some will try, it will probably falter quickly and exit the ecosystem.
CH: Many people are also talking about the possibility of an AI bubble, at least in terms of the financial arrangements between the largest companies in the space and the wider stock market. Should that come to pass, do you feel like the healthcare industry would be outside of the blast radius?
BM: For me, most of that talk is around chipsets, data centers and the amount of investment that's going in, and the amount of return that's going to happen there.
Look, let's assume there is an AI bubble, which means people have over-invested and over-assumed what's going to happen. I think healthcare isn't as impacted, and the reason why is it's such a small segment today within the use model.
When you look at the overall usage of AI, the bulk of that is going to be in the consumer world. I don't think healthcare has advanced today in terms of relying on it.
If you were to tell me we have to go back and not use AI, I'm gonna say, okay, we can go back to calling it an algorithm, whether it's on a local chip or in the cloud. But really, we're not going to be as impacted… At the end of the day, we're still going to be able to hopefully deliver the outcomes, but maybe a little bit slower, right?
Because I don't think this ecosystem is going to be a healthcare company delivering it on its own. I think it will require partnerships from some of the big boys to come in and help out with some of the overall models—the Googles and Amazons of the world. So it may slow us down a bit, but I don't think it'll have a significant impact on what we're trying to achieve.
CH: When it comes to BD connected care—you talked a little bit about the industry seeing more SaaS models—will it be a big transition for BD to grasp that digital service provider mantle, instead of simply delivering refrigerator-sized Pyxis machines?
BM: So the true question underneath what you're asking is, can BD do a solution sell versus a product sell? That's the first step to capturing that SaaS model.
Look, this will be a journey. This isn't something that will happen overnight for us, but I think we're definitely invested in that sales excellence piece, and bringing in talent that can actually sell a solution—and I think our customers are ready for it. That's what's a little bit exciting.
You know sometimes you're building something that's pushing against a wall, or pushing against the ecosystem, but I think the ecosystem that we’re playing in is ready for it. And over the next few years, we'll probably get better at solution selling, and the ecosystem will be more welcoming.
But the timing is right, I would say, for BD to go after it this way. I think, based on our current market share, and the category leading products we have, it puts us in the right position to start to add value across what we deliver—and that value, I think, will be realized by increasing our overall addressable market by being in a SaaS model.
CH: And for this roadmap, does it include—we're talking a lot about almost automating which drugs go into which patients—do you foresee stops along the way at the FDA, with regulatory submissions for any one of these programs?
BM: Absolutely, and that's not new to us. We have probably three clearances already that have to do with AI, especially on the automated pharmacy management side of things.
But we're splitting this into two categories. There will be clinical outcomes we're trying to achieve—reduction of length of stay, bounce backs to the ICU, bounce backs from home back to the hospital—all these types of outcomes in the ecosystem we're going to go after, and there'll be regulatory clearances associated with them.
Then there's the efficiency play, which is another thing—and the efficiency play is massive. The dollar amounts we can save hospitals by predicting what's going to happen with their medication management is huge.
And I think of both of those, this will probably be less intensive from a regulatory standpoint—I'm making assumptions here, but maybe less of a regulatory burden to get through, even if they do have to go through a 510(k) clearance. So I think it's a journey, and I think we already have the muscle at BD to go after it from a regulatory standpoint.
We’re super excited about what BD can deliver during a time where many hospitals are struggling—they're being pressured to manage the change from volume-based to value-based care, and caring for those patients all the way into their homes.
I think a lot of medtech companies haven't been able to fulfill all those needs as that transition occurred. But I can tell you today, in talking to some of the top hospitals we deal with, as soon as we mentioned, the ability to start looking at data across the different portfolios we have—they’ve started to come up with the algorithms and solutions that are possible way before I say anything. ‘Oh, you could do this, you could do that.’
And that's a good sign for us. That means they're ready for it, and they're excited about it, and they see the impact it will have. So when it comes to the next phase of developing it and getting it out there on the market, we're going to have true partners that will help us get there.
Editor’s note: This transcript has been edited for length and clarity.