Welcome to Real Talk on AI and I care with your host, Dr. Scott Morris. And I'm co host, Dr. Rehan Ahmed.
In each episode, we discuss current news, adding a little AI education, and debate innovative AI tools and topics that will change the industry. Welcome back to another episode of AI and I care. Scott, how have you been doing? Good.
It's been interesting. You know, always pre-holiday. It's just crazy busy of stuff to do and things you want to do and things that you should do and things that you never get done. So it's all good.
I'm looking forward to this one though. We, for the audience here in our pre-talk, are kind of our prep meeting. I think we got some interesting things for you guys today. So for all of you, and thanks for all of your feedback.
I've gotten a lot of feedback the last couple of weeks that, you know, folks who are like, man, I listen to you guys every week and it's how I keep up on what's going on. So I appreciate that. Rehan appreciates it. Please let us know how you're doing.
We'll leave, we'll give you our emails at the end. And we love to hear ideas and feedback. So thank you for all of that. Rehan, tell us, you got some interesting news of the week.
Yeah, yeah. Come here. This is a piece of news that I think speaks to a theme that we often talk about on this podcast, which is the eye as a window to our health, right? It's systemic issues.
And so this is about a company called Harmonize. It's based, it's AI platform company based out of Bethesda, Maryland. And they folk, they acquired a company called IFocus Health from California. And so it's all what does that do with anything with AI and I care.
But the reason it does, this company Harmonize, that's H-A-R-M-O-N-I-E-I-S, they built an AI model with like 14 million records. And what it does, it essentially has learned how human eye movements correlate, get the Scott with emotional and cognitive states. And so they can figure out what? Yeah, they can figure out a lot of someone just by watching their eyes, which is interesting.
But here's the, the thing about what that means to psychology and society, that's crazy. And I think that it's an interesting beachhead that they're going after. It's ADHD treatment. So they acquired a company called IFocus, which has very specific knowledge on the relationship between when a kid or an adult starts ADHD medication.
So there's a big problem, like how do you know it's working. And it's usually subjected, you know, mom or dad or the parent, they thought it's questionnaires, they thought forms, it's kind of very slow and prone to bias. So what IFocus does is that they use a webcam and they track patients' eyes while they're doing a simple task. And the science is basically that you're focused on attention, basically, you know, executive function, prefrontal cortex type of things.
They are the ones that control your eye movements or what you're focused on, sort of makes sense. And so the AI can see loss of fixation stability or in suppressing sort of a reflexive eye shift. Like, you know, think about the stochastic movements. So it's so funny.
It's so funny, and you talk about this because I did an FDA study. I was not the lead, I was the second on it. About five or six years ago, that was looking at just this topic about ADHD and fixation disparity. And we found a third of all people who were diagnosed with ADHD had fixation disparity and you treated with prism and their ADHD magically went away.
The other two thirds didn't, and we could never figure out what's going on with that. Maybe these guys have figured out the rest of the story. Maybe they had to really cool. This is something I would love to have them on.
So I think we should definitely dive more into this story. But I think what's relevant for our audience is you get another way. I mean, often have to think about the eyes of the window to systemic health, we're thinking reddened, right? And I've given sort of multiple talks where it's okay.
And I'm telling you, this is not, it's not just a reddened story. Well, the reddened can give us tons of information on Alzheimer's, neurodegenerative issues, cardiovascular issues of course. But there are other things. There's eye movements, there's eyelid movements, there's anterior segment.
So, you know, the great thing about the eye, it is right there for us to see. And it gives us so much information here. It's giving us more neurologic information on ADHD medications. It's all about you talking instead of biology, the window to biology, it might be the window to psychology.
Oh, yeah. That's crazy. So super interesting. This news just broke yesterday.
So we're taking this December, loves it, broke December 10th. And so we look forward to writing more about a journal at AINICare.com. Rayon, that might be an awesome one for the innovators series, because that's really cool. And that really is the next step of AI, right?
I mean, like you said, it's not just not just a culomax and not what it's going to do in terms of practice workflow. But maybe we're going to be able to help a whole bunch of people who just struggle to learn. That's amazing. That's a great story, Rayon.
So it's great. And so for our optometrist and ophthalmologist or listening are providers, this is yet another way that you can theoretically interact with your patients, right? There may be a future where patients are coming in and you're not, you're diagnosing them and talking about neurodegenerative ADHD as being that first, like, hey, your eye movements, there's something potentially here, guys. And so yeah, super excited to talk about this much more later on.
I got to dig into that and learn more about that. That's pretty cool. All right, well, so the audience, you know, we always do our knowledge by, and today, I, Rayon, we're talking about this. I've had like three or four people in the last week talking to me about, well, Scott, what's an agent?
And what is it going to take over? Like, what's it going to be a ton of this? And just, you know, do all these things without having us to ask it? And I was like, wow, I, you know, that's a big subject.
And so, Rayon and I decided, maybe we're going to talk about one in the time's agent is. And then maybe for our fireside, we're going to cover, where does that fit into the grand spectrum of agents? And is that factor fiction? So let's kind of roll out a little bit about what's an autonomous agent?
Well, you know, you guys all heard obviously, AI, automation and all kind of stuff. But the, the term we're hearing a lot about in terms of clinical efficiency is, are we going to have an autonomous agent? You know, I, you can kind of, for all of you guys are listening to their I care professionals, thinking about the difference between a manual foropter and kind of a skilled technician. One requires you to turn every dial yourself.
That's the foropter. And the other understands, hey, you know, I know what you need. You need a refraction. And you need an outcome that's a good visual acuity.
And they're going to go get it for you and then present it to you, right? And so, you know, Rayon, in the optometric industry for a long time, I've Tom Kee said, well, we're going to do this ourselves. And I think a lot of the big practices and I come from more of the ophthalmology side of the world for the first big chunk of my career is, my way at staff, they could do that. And they were amazing at it, right?
And so, I want you to think about they're working somewhat autonomously to get the job done that we want to get done. And if you think, so when we talk about autonomous agents, you know, like, let's maybe talk about this as kind of like your core piece of it. It's a software program that doesn't just follow a strict algorithm of instructions. It starts using, and this is where artificial intelligence comes in, it starts perceiving its environment.
It starts reasoning like, well, how do I solve this problem? And it takes some education, some training, and lots of, you know, we've always talked about data, data, data, data. But then it says, I'm coming to learn from my previous things. I'm going to learn how to solve this problem.
And I'm going to take action to get to the goal. So now instead of a very set algorithm, we're starting to say, here's our goal. How do we get there? And so, you know, and I think you can break that down into two things.
There's like our standard automation, which is what we kind of think right now, which is the script. You know, if you go onto a website and you click schedule and appointment, then a calendar opens, and it brings you down a very strict algorithm of how you get there. Whereas an autonomous agent would say, hey, Michael's filled the schedule, and they say, okay, there's a cancellation. Let me look at the wait list, and then identify who would be the next eligible or high priority patient to get in, and then send some text and then bookstay appointment.
And your human staff says, oh, look, somebody gets scheduled. You know, it's kind of taking some of the scribes that we talk about and some of the scheduling we talk about, and some of the, you know, and that's what the future of the digital employee may looks like. And why is that going to matter? Well, you know, think about them being able to do triage, right?
And say, hey, not just let's follow some algorithm, but let's actually, when a patient describes flashes and floaters, they say, oh, this is urgent, I need to look for what the next emergency slide is, and I'm going to ask you a bunch of questions, but while I'm doing that, I'm also going to see when can I get you in? Without having your staff review emails or text messages or, you know, whatever communication system you are using, your staff has to view, now take your staff out of the picture. Scribes, you know, when we talk about scribes, we've had to entire things on scribes and that they're their own spot from an autonomous perspective. And then you get into other types of agents which we'll get into here in a little bit, but I think of it in tournament agent as a just very goal-oriented software.
It lets you delegate outcomes and says, you figure out how to get there to give me what I want. I don't know. Now, you know, so that's kind of the knowledge by the day, but I think it's going to lend into our conversation, you know, Rayhahn, you and I get asked this and we talk about this a lot, is where is this? Is this fact?
Like is autonomous agents here? Is it fiction? Is it like, that's something they talk about but it's never really going to happen? Or is this really just the future that hasn't happened yet?
Yeah. I think this is a work in progress. We are still there and obviously we'll talk much more about this, but we are at a point now where these are being deployed for task-specific type of things. So many clinics are using agents for scheduling.
I think that's becoming a much more of a common thing and agents for answering the phone calls or having some specific knowledge for the patient or the clinic. I think that's definitely here. Now, I think you described something that's super interesting, kind of like an agent of agent systems where there's a higher order of agent. So you have someone, some agent at the top and agents at the bottom.
I think we're not there yet. I've not seen any clinics deploy that. Don't know when that's going to happen. And one of my, I think I say this often is like, overestimates going to happen in the short term and underestimates the long term.
I think that's going to take some time, but when it hits us, it'll be, it'll be, it'll be coming fast. Because the issue is we, I don't think we've perfected a lot of these low-level agents still are still human in the loops and they need a lot, right? Even with scheduling. But we're still training them, right?
And when we talk about training and training. We're getting it still. They're learning. But I love how you said they're low-level agents, right?
So they're doing somewhat specific task-based problems. Very much like gender to AI was doing five years ago. Right? If you go back five years ago and think about when GPD one first came out or whatever, and it could do very, it was kind of limited, right?
And now it can do crazy stuff. And at least in what we think of crazy and in five years, we're going to look back and go, this was so, you know, set of a toddler was a teenager. But I think you said it great. You know, there's low-level ones.
They're going to learn how to do these tasks real well. And then there's going to be a mid-level agent that says, well, let me take the scheduling data and the HPI data. And maybe the P-H-R data and maybe look at what workflows look like that day. So there's four different low-level agents.
And you're going to have a middle-level agent go, okay, let me, let me put all that together and say, when's the next available time to get that person in that's not going to slow down workflow? And then maybe you're going to have a couple of those, maybe one that does a scheduling, you know, that's kind of in charge of workflow efficiency. And maybe you're going to have one that helps with clinical decision-making. And maybe you're going to have one that helps with the backside, you know, operations, the business.
And then you have a third level, a third tier. Let's call it your, instead of your local manager, you have your office manager, your your practice administrator, agent, you're saying, well, let me take what my middle managers are doing and put all that together. But that's going to take time. Right now.
Because at the end of the day, what is an agent? Basically an agent is, is any type of software, it's AI software that is not waiting for you to do the next step. It will independently take the next step on its own. And so what that means is that a lot of these tools, systems, they're, well, we're going to have to build trust in them.
I mean, it's not going to happen overnight. We're going to have, they're going to have to prove to us that they can do these quote unquote low level skills. And that's what we talked about last podcast, right? We talked about X AI and, you know, the big pizza that's going to be trust.
I love that you kind of brought that back to that discussion is, do we trust it? I don't know. Yeah, yeah. I like, prove your worth over the next couple of months and years and show, show the team, the clinic, the doctor that you're able to do.
This is a way that's great for patients and great for them and great for the, the clinic and so I think that will take time. What is that any different, but Ray, is that any different than when you hire a new human staff member? You're kind of like, do I, I'm not going to throw you in and have you start doing all this stuff day one? You kind of kind of build up my trust ago.
Do I, do I trust what you're doing? Like did you do the right things? Did you ask the right questions? Did, did you do the test in the right way?
And when, when we hire somebody that's, we hire a biologic human, right? We, we have to learn to trust it. You have to learn to trust now. I don't think this is any different.
I mean, I just think it's, it's a digital system instead of a biological system. Yeah, sure. That's good. And you know, they both have the same problem of being quote, a black box, right?
You can't like, peer into someone's brains and say, Hey, what? I'm, why did you, you could ask them, you could also ask the AI system and try to figure out what happened. So yeah, I think there are some similar. Of course, very meaningful and very important differences.
No doubt about that. But that's a good point. And, and I loved actually that brings up a really interesting point. When you're comparing an AI system, when it makes a mistake, or it's 99% accurate on the transcription, but not 100%.
I always find people say, Oh, it's not perfect. It made a mistake. It did this and like, well, you know, we'll compare that to the human. Is it better than what the current, yeah, gold, gold standard is right now.
So I think that's a, that's actually really great point. You, we have to have fair benchmarks, comparing different type of solutions. And so sometimes it's, we're putting the new AI system, especially, and I saw him to a founder of a company that does pre-authorism. And he said, we're 99% accurate.
And clinics are saying, well, why are you not 100% accurate? And so, well, and then he said, well, are you, are you, are you or I would love if my, and I love my team, they do an amazing job. I'd love to say we're 60% accurate. I think that'd be a huge positive, right?
So oftentimes, we're not getting data on our current solutions, because up till now, we actually didn't have another alternative to comparing it. So I think when you're, when you're thinking about adopting a new AI system for your client, for anything really, think about what you're currently doing. And then you have to have a way to measure the efficiencies and the outcomes and then be able to compare the Bay-I system, right? Because you can't just say, well, the AIS is worse or making these mistakes.
If you don't know how bad your current one is right now. I think we don't want to know how bad our current system is. That's right. I really, I sometimes go, maybe it's better I don't know, right?
I mean, because that's what I'm saying. I'm not saying that, I'm not saying that, I'm saying that, I don't know, right? I mean, because then it, then it seems like it's working well. I love that.
I love that. You know, I, I'll tell you another agent I'm really looking forward to happening. And I've, we're just not there yet. This is still a, I, I would put it in the fiction category, but I think it's not far away from back.
I think one of the coolest agents we're going to have is a learning agent. I, a learning agent for us as providers for our, for our staff, for our patients. I think when you get a learning agent that takes over some of the educational pieces that suck up so much of our time and granted, they're still very, very important. But I think in terms of an agent that can do pre-education and can do live time in office education and.
And I'm, I'm, I'm, I'm, I'm, I'm, I knew, for a new staff number. Yeah. All of that. I was gonna talk about this and here's myself getting myself in trouble again, but you know, could you, could you, could you, could you, could you, could you, can you, can you, can you, you, can you, you, can you.
I think we're probably three to four years away from those. And some of our colleagues are working on these things right now. I think that might even be more important. For the flow for the low and business of eye care that even like scheduling, you know, clearing house and all that kind of stuff.
I think that's gonna be a big piece of the puzzle too. Yeah. Teach us what to look at. Like here's a spot of retinol thickening and this is what this means and we're gonna be like, oh, I know as an even surfer if I really saw that was retinol thickening but thanks for letting me know and what do we do with this, right?
So I think that's gonna be really cool. I think that, but I like kind of what you said, I kind of go back to your low level and then we take mid-level and higher level. I think, you know, this isn't, I think this is gonna be a philosophical shift, Rayhan, is that right now we hire people and we think about buying or subscribing to software. I think we're gonna change that philosophy.
It's not gonna be about subscribing. Here's a philosophical challenge. I think we're not gonna be buying or subscribing to software. We're gonna be hiring a team of digital agents.
Big difference, right? I've said, oh, this is just software. Now this is a digital interface that I interface and talk with all day long and my staff interface, my human staff interface is all day long. I don't have a crystal ball but I don't think we're five years away from that.
I don't think it's that long but I also don't think it's next year. It is hard to make a prediction when that's happening. I can tell, I got a spam call today and it was just amazingly good. And I knew they're using one of these next generation voice models because it was, it sounded like a real person and it's response time was really good but there are certain repetitions are you not a huge, you know, like you could tell when it's like the voice bots a year ago, you could tell.
Well, we can still tell barely. It was good. Yeah, I think it's getting harder and harder to tell. It's getting harder and harder and now even the deep phase.
So yeah, I agree, the pace of change and like you mentioned, Nana Banana from Google Gemini that these things are getting very good very quickly. I mean, that was just about a couple of weeks ago and I mean, completely revolutionized the way we make educational videos that are office completely flipped the switch in one week. Like literally my marketing person that does that, she's like, yeah, I can start turning out what used to take me two weeks. I can turn out in an hour.
Yeah, like wow, that's using, I mean, I don't want to call Gemini truly an agent but and Nana Banana or whatever we call it. I don't know if it's truly an agent but boy, it just changed workflow a whole bunch overnight. Yeah, so and we're still, I think we're still in its infancy. So if we look at agents, I think it's gonna be a blur between gendered of AI and agents and algorithms and I think it's gonna be an integration between digital staff and biological staff.
I can't wait, it's gonna be fascinating. You're coming under the challenges like you said, I mean, there's definitely gonna be challenges and know it out about it. But I think we have to, as you said, put things in context of, are we willing to teach digital like we're willing to teach biological? Yeah, exactly.
And I think our expectations have to be tempered to what we had in our old, biological only clinics. Sounds weird to say that, right? But anyway, we're right. It's always good to talk to you.
Hopefully for the audience, you guys learned a little bit. We made it think a little bit and fireside, go, huh, what would the future look like? And if I had to hire a digital agent instead of a biological staff member, where would I, what would be my next hire? And then am I willing to teach digital like I'm willing to teach biological?
Something to think about over the long holiday weekend, over the long holiday couple of weeks here. Yeah, yeah, exactly right. Well, everyone have a wonderful holiday, few weeks. We will be back in 2026 and looking forward to continuing strong with our journal and podcast.
I appreciate everyone's attention and time and feedback, Scott, and I appreciate you. I appreciate you too, my friend. I look forward to the new year working with you. And for the audience, we have a, we've kind of already started designing our calendar for next year.
We're going to get deeper and some of this stuff will always do our knowledge bikes. We think it's important, but our goal is to help you learn more about how to improve what you do. Everybody have a good holiday. You've been listening to Real Talk, an AI and Icare, your weekly podcast to keep you informed about AI technologies revolutionizing Icare.