This transcription is provided by artificial intelligence. We believe in technology but understand that even the smartest robots can sometimes get speech recognition wrong.

[00:00:00] This episode is brought to you by Artisight. In today's healthcare environment where clinician burnout is at an all time high, Artisight's smart hospital platform delivers measurable results. 94 percent nursing satisfaction, 26 percent reduction in overtime, and 45 percent lower turnover.

Their AI powered solution streamlines clinical workflow without burdening your staff, creating what one healthcare CIO calls nurse satisfaction through the roof. Learn how Artisight can transform your clinical operations at ThisWeekHealth. com slash Artisight.

Welcome to This Week Health. My name is Bill Russell. I'm a former CIO for a 16 hospital system and creator of This Week Health, where we are dedicated to transforming healthcare, one connection at a time.

Now, onto our interview

(Interview 1) All right. Here we are from Vive 2025 in Chilly, [00:01:00] Nashville, Tennessee. I'm here with Andrew Gostine, the doctor. Andrew, every time I $introduce you, I don't put the doctor in front of it. CEO of Artisight. And I'm excited for this conversation. There's a lot going on here. I was just interviewing somebody and I said, what are you doing in the patient room?

And they said we just signed on with Artisight. So you guys have a lot of momentum going on right now.

Yeah, we had a great 2024, 2025. We've already doubled revenue in the first month and a half of the year. So things are really taking off for us.

So let's talk about this for starters.

What's happening in the hardware world? We'll move through the software AI models because I think that's what

differentiates you guys. But let's start with hardware. A lot of the customers were coming to us, historically we had this hardware agnostic approach where we could use any camera, any speaker.

And for a lot of customers, they like that about us. But some customers don't have cameras, speakers, any of those things in the hospital and want a very easy to deploy solution. So we developed our own [00:02:00] camera, which is actually two cameras, eight microphones, two speakers, an alarm, our indoor positioning system connects with Ethernet, Wi Fi, 5G cellular connections to really solve all possible permutations of going into a hospital.

All on this? All on a single device. With a full NVIDIA GPU inside it, so we need no cloud compute to do any of the processing for any of our algorithms.

So you don't have to stream, you can actually do the inference engine is in the box. So

for all of our algorithms, we have about 32 different patient room algorithms right now.

They all live inside that device. Nothing ever leaves the patient room. That's pretty amazing. Tell

me a little bit about the camera. Cause you do work with a lot of different ones, but my guess is you probably picked one here that. That gives them the ability to do some things.

A camera, to an extent, is just a camera. It's more about the number of cameras. And so if you look closely, we actually put two cameras in it. That way, a physician who wants to remote into a room and do a [00:03:00] pupil exam can do that with a very high powered camera, while the AI still watches the second camera to make sure the patient doesn't get out of bed and fall.

Or if a nurse happens to be doing a patient . Turn at the same time, you're never going to have that blind spot. So it's really two cameras is more important than the quality of a single camera.

of the things that's interesting about this. Having been a CIO and done construction projects and upgrades, we have to shut those rooms down when we do those kinds of things.

That's one of the big initiatives that we thought about when we were developing this device is to make it a lot easier to install into the patient rooms. About three different clamps that can actually just clamp it onto the TV or hook onto the mount for the TV. That allows you to install this securely without drilling a hole into a wall.

Now from the infection control prevention standards, if you drill a hole in the wall, the patient can't be there. If you're clipping this onto a patient room, TV, a patient can actually be in the room or [00:04:00] just in between when they're going to get a CAT scan, you can run in quickly and hook it on the TV. I would imagine that enables a health system to deploy this a lot quicker.

From our perspective, we want to make sure cameras can get installed as fast as possible. Helps our revenue growth, but also, from the hospital's perspective, they don't want to pay deployment teams for six months when they can do it in two weeks. So what do I need? I need power and Ethernet, essentially?

It's a single cable, so it's all PoE powered. So it does take a PoE power supply, so you have to make sure you have the appropriate power. If you don't have that in your network infrastructure, it can use a 120 volt power support, source. But you gotta remember, this is a full NVIDIA GPU inside it. So this is taking power here to run all of the processing.

Why do we have a TV in front of us?

It's not enough to just do the camera, right? With remote nursing, with all of these new use cases and care team models coming out. It's about bi directional video. And so if you want to do bi directional video, you have to have control of the TV. We have only ever [00:05:00] found one client that would have been happy to install two TVs in the patient room.

Everyone else wants one. Everyone else in this space uses HDMI CEC to control the TVs, to change the inputs back and forth between patient content and teleconsult. That's fine if you're doing a pilot. It will work in small volumes, but Northwestern Medicine. Does 5, 000 two way video calls a day now. CEC might work 98, 99 percent of the time.

That's between 50 and 100 help desk tickets coming into IT every day because the TVs didn't respond with that consumer grade technology. So we actually developed an entire TV OS with Samsung, with LG, and integrate securely with PDI TVs as well, so that we can take API level control of the TV. To get that five nines of responsiveness to eliminate issues with TVs not turning on, not changing [00:06:00] channels, not changing volume.

So what you're seeing here now is because we did all of that work, it comes with a lot of other cool benefits. Putting different sidebars on the TV. Putting patient entertainment on the TV. Clearing the patient's Netflix account the second the discharge order goes in. There's all sorts of automations you can build once you control the window.

So help me to understand that. I'm buying a standard LG, I'm not buying a specific LG TV. You just have the OS that gets installed on the TV.

It's specific hospital grade TVs. It can't be anyone. It has to be their hospitality grade TVs. But anything going back about five, six years, for both LG and Samsung, we now do all the firmware patching for these devices.

We take full maintenance control of the TVs for the hospital staff. That's pretty amazing.

Unbelievable. Talk to me about AI. The conversations we've had over the years, I've, first of all, I've learned a ton just from the conversations, the distinction I always make with AI, artificial intelligence should learn, and [00:07:00] there's so much artificial intelligence that's just AI washing.

It's like, we have AI or AI is in the name or whatever. It's like we do it. And essentially all they are is glorified algorithms. I love our conversations. Talk to me a little bit about what you guys are doing in the area of AI and integrating it with workflow and really supporting the clinicians.

So about now we have roughly 17, 000 patient rooms and about 1, 000 operating rooms online. We'll probably triple that this year. So with that footprint across healthcare in the United States, it really represents a very diverse environment for us to train algorithms. You cannot develop algorithms to a single problem in a lab, try to turn it on in the hospital.

It never works with the hospital workflows, they never catch all of the edge cases. If you want a real AI vendor, they have to be able to explain to you how they train these algorithms, and the algorithms continuously learn in your [00:08:00] environment. That's what makes it work for the doctors and nurses. That's what makes us want to keep it on and not get a bunch of false alarms.

So Artisight benefits tremendously from all of the work coming out of the hyperscalers now to open source major algorithms. This is really the year of multi modal models. And by that I mean it's not just computer vision or not just text analysis or generation by a large language model. It is really a foundational model that can ingest multiple different data sources across different modalities like vision, audio, text, or indoor positioning information.

And that's what allows us to continuously train algorithms and update them, even as workflows in the hospitals change. That sounds very nebulous, a great example. We've now released, because we have all of this hardware in the hospital, and we have control of the TVs, full ambient nursing documentation.

A nurse walks into a room and says, I'm turning the patient to their right side. That will allow us to document that for the nurse [00:09:00] inside Epic, inside Oracle Health, inside Meditech. But it also teaches the computer vision system. With a GPU inside the device, you just witnessed a patient being turned.

That audio gets converted to text. The text then teaches, through a large language model, a computer vision algorithm. The computer vision algorithm now understands what it looks like when a nurse turns a patient to their right side. So every time a nurse or a physician or a respiratory therapist or occupational therapist goes into a room and vocalizes what they're doing, it's simultaneously teaching our computer vision to monitor that task, document that task, and offload that responsibility from the clinicians.

Practicing in a hospital, used to be like, you hired a physician and they got used to Epic and then they went to another hospital and was on something else, they'd be like, Oh, I learned this or whatever it happens to be. I could see that happening now.

What you're describing is going [00:10:00] to be a very distinctive experience for the clinicians. If they're practicing here and then they go over here and they don't have it, they're going to be like, Are you kidding? I have to go up to the keyboard? going to be an old model.

You touch on a kind of famous quote we have at Artisight we really do envision a future where there are no keyboards and no mice in the hospital.

It really could be done as a fully ambient experience. And to your point, we do see turnover rates at our client sites typically cut in half within the first six months. People don't want to leave this technology for the same reason that they're in love with ambient documentation. There's very few solutions out there that ask doctors and nurses to do fewer of the tasks they don't enjoy doing.

And if you take those away from them, you can solve a whole bunch of healthcare problems. So

as it scales, you're going to get more information. It's going to get smarter. It's going to be able to recognize more and more things.

By the end of this year, we'll 67 unique events just in the patient [00:11:00] rooms. In the surgical space, within about the next four months, we'll release algorithms that can watch all endoscopic video and fully document for the surgeon the entire operative note.

Just by watching the surgical video in real time with, again, a GPU in that environment. All of the manual documentation, task tracking, notifications will move to Ambient in the future. And because, like you said, this essentially becomes a positive feedback loop. The more we can provide the hospitals, the more hospitals like it, the more it gets accepted by the doctors and nurses, the more training data, the better and more sophisticated these algorithms get, until we're essentially solving, theoretically, every problem.

Fantastic.

Andrew, always great to catch up with you. Thanks for coming by.

Thanks

for listening to this Interview in Action episode. If you found value in this, share it with a peer. It's a great chance to discuss and in some cases start a mentoring relationship. One way you can support the show is to subscribe [00:12:00] and leave us a rating. If you could do that would be great, thanks for listening. That's all for now.