Welcome everyone to a special episode of the Innovator podcast series called AI in clinical trials. And I'm very lucky to have three amazing guests here to talk about the way AI is reshaping clinical trials and eye care. Arshad needs no introduction. Dr.
Kanani is a director of clinical research at Sierra Eye Associates. As Ben, as you were saying in our pre-meeting, he is the best-dressed person and eye care. Best ophthalmologist on LinkedIn. I look at his post basically for what he's wearing.
He is not in the who's who of retina specialists. If there's an important retinocle trial, he's definitely on it and he needs no further introduction. Ben is CEO and co-founder at Amros based in the Bay Area. It is a precision intelligence company devoted to ophthalmology.
So very lucky to have Rom Yal and Munchili co-founder and CEO at Tilda Research. Is it Tilda Research or Tilda Bio Rom? Tilda Research. OK.
So very lucky to have Rom Yal and Munchili founder of Tilda Research, also based in the Bay Area and a company that is really talked about having AI teammates in clinical trials. So Arshad, I'm saying we don't have a ton of time, so I want to skip this start and we're going to jump right in our team. So tell us a little bit about you've been in this space a long time and you have probably seen so many challenges and problems and follow the next with clinical trials from recruitment to making sure endpoints are hit, et cetera. Where do you feel are sort of the biggest challenges for clinical trials, specifically in ophthalmology, and where do you think AI is best positioned to address those?
Around, first of all, thanks for having me here and thanks for all the kind words. And doing clinical trials for a long time, you can see all the gray. That's all from clinical trials. But I'll tell you, this is the most exciting times for clinical trials, not because we have innovation, but actually we have AI.
And I'll tell you, AI will play a key role at all the different levels that we face challenges. So I look at three different buckets when I look at clinical trials. Number one is the site. So what can we do at a site level?
So we have continuous changes in staff. It's very hard to keep staff. We have to retrain the new assistance. They come in for six months, 12 months and they go to med school or other careers.
So we have like this staffing issue that happens at every site. We'll be lucky that we have some key pillars that are staying with us for 15, 20, 25 years, but we need a lot more help than just them. So I think AI having AI teammates, and I'm sure we're gonna get into it, what Ron does, it's very exciting that we can use teammates to help us do regulatory, do EDC, do budgets, do accounting. So that's an unlimited opportunity from a site level.
The second challenge, a site level that Ben and Ambrose and other companies can help us is finding the right patients for clinical trials. We see a lot of patients, for example, geographic atrophy, right? And we see, oh yeah, we're gonna have a trial comment, we'll call you. And then six months later, we're like, oh my God, where are those patients that we had?
And did we have a sheet that somebody updated the EHR and who was that patient? Well, AI can actually, mind the data, can look at the lesions, can fit the inclusion, exclusion criteria, and help me lower my screen fail and find patients that are, now that there, that it's gonna be good candidates for clinical trials. So those are the two things at the site level. The other two things obviously are sponsor level and CRO level that AI can help them.
But from my site level, I think AI is gonna revolutionize what we are doing in retinoclinic clinical trials. Ben, can you take that on here a little bit? It's obviously a lot that can be done to the site level and how third parties can come in and really help the site and then also help the sponsor. So speak a little bit about what Dachanani was talking about and some of the challenges and where Amorous really helps us problems that he was talking about.
Yeah, yeah. Thank you for the invitation and opportunity. Dr. Konani already mentioned the challenge and one of the things that we do is we work with both the sponsors as well as the practices, like what we do is we connect five different data sets under the same umbrella to be able to do these analysis that for patient identification matching with precisely to the existing study.
And those are very quickly EHR claims data, provided notes which plays a huge role and we have large language model to understand the intent of the content. Imaging, huge, we believe it's 70, 75%, especially on the post data segment plays a huge role so we have biomarker and alles biomarkers that we analyze that helps the accuracy as well. The fifth one is genetics when it's available. So all of them under the same umbrella, it gives us an ability to the precision intelligence which is one of them is for the patient recruitment as well as helping the site selection on the CRO site and on the sponsor side allows us to run criteria as a simulation to help them optimize the criteria as well.
So we're in the middle, we work with the practices as well as the sponsors and CROs. Fantastic. And Ramen at Tilda, so I went to your website and I love this term AI as your teammate and that's something we talk a lot about in our podcast at AI and I care. How it's not replacing you, it's just sort of augmenting abilities and so I love the word teammate.
And so what do you guys at Tilda, what does that mean to you and sort of like giving you some examples day to day, how does that actually, how does it an AI be as a teammate of mine in the context of the protocol? Yeah, absolutely, Ramen. And thanks for having me excited to be here with the whole Dr. Kanani and Ben.
I think from my perspective, you know, I've, I don't come from a deep medical background. You guys know that I'm more of a technologist. I've been in the traditional machine learning and AI space for quite some time for several companies in the space. But what is unique in terms of what we're seeing today and why I think I would have to double down on Dr.
Kanani's statement that this is the most exciting time to be in clinical research, like Parnum, right? I really do believe in that. And part of the reason about that is historically we have never had a type of technology which can massively increase the bandwidth or the throughput of an infrastructure like the clinical trial infrastructure, right? So you've got this very valuable resource which is really constrained.
And I've spent a lot of time understanding the site business, sponsor business, you know, I've also co-founded a company called Lex and Fire Cutilda, which we got recently sold to Roche. So I've spent a little bit of time understanding the market from different angles. And I think one of the key words I've always come back to thinking about is today the most premier infrastructure, if you will, the site infrastructure or even CRO infrastructure depends on the attention of your key staff. That is the last, like that is probably the biggest indicator of how well you're gonna do.
You go to a site like Dr. Kanani's, look at his top five people. I'm sure they've been there with him for quite some time. I don't know them.
I'm just taking a while guess, but I'm quite sure about that, right? I go to another physician's office. It's such a great, like indicator of success in the traditional infrastructure play. And what's happening right now is, you know, for many reasons over the years, retention has actually dramatically formed off the cliff in the Quenco trial space.
You talk to CROs, you talk to sponsors, you talk to sites, this is a major issue. And I think for us, it's always been about how do we expand the bandwidth? The shocks on the goal is the only thing I would say, I'll always do a bit care about as a sort of your meta metric to see how much of an impact we can make. And for me, that's what it is, right?
How do you bring a teammate who can help with that, with that retention overall? Now that could be better work-life balance to the staff, better, you know, augmenting their ability to do more with less. So there's several ways you can cut out, but enabling technologies need to comment whatever those might be, right? Like agentic solutions or teammates or whatever you wanna call it.
Now I'll spend quite a bit of time in building these, you know, we're about 100 plus AI engineers and PhD researchers that I told that. And we've spent the last four something plus years essentially thinking about this problem and training these AI models to get to where we are. And one of the interesting things which we have done, or at least started to realize about two and a half years ago is on the spectrum of where AI is today, we are moving away from the hyper growth and the conversational AI, which is like chat GPT and sort of your multi-model chat, right? Like chat box to an agentic AI model.
And 2025 is definitely the year of the agent model. We've made several breakthroughs on the model front. There's been reasoning, there's two calling. There's so many interesting things which have happened in terms of breakthroughs on the technology side, which then enable companies like ourselves to build these really aligned AI teammates, which feel really natural, which can do a lot of the work which you normally do manually.
And there's a lot of interesting technologies where the future looks a little bit interesting actually. So I think for us, that's what it is. Like, how do you build these AI teammates in the pillars of clean cooperation? We focused on just the problems of pure operations.
And so, typically those would involve data management and data monitoring, regulatory management, all the TNF type of workflows, finance operations and site and study management. So all your startups, closeouts, maintenance, and related to all that work, right? This is like bread and butter of where the administrative work has turned into the clinical research industry. I'm sure if you talk to eminent researchers like Dr.
Panani, the fun part is not the research administrative part. The fun part is to work with new drugs, see the medicine move forward and all that stuff, right? So I think I come from that perspective. I'm just saying like, you know, let's go build these AI teammates which basically obviate the entire admin study work.
We don't want to change the world from an admin perspective. The process exists for a reason. Now we just need to find ways to like reduce that burden on us as humans. So we're not focused on the admin work, we're focused on the fun part.
And I think for the very first time, we're showing that to our customers, we're showing that to the world that this is doable. These are highly aligned AI teammates that we built. And you know, it's been awesome. Like, just grab so much growth in the last couple of quarters.
That's very exciting to see where things are going into the next couple of months and course. Fantastic. Dr. Panani, I know you're not one to get into any BS or anything like that.
And every time I talk about... Tell me AI, we know me well. Yeah, and so pretend Ben and Rahm are not here for a second. And everyone talks about the AI hype cycle.
You know, it's like, where are we? Like, there's so much hype. In fact, we get a ton of feedback here at AI and iCare. And some of it is...
I mean, I'm totally honest, it's like, you know, what about the patient, doctor, bonds? And, you know, AI is going to take all our jobs. There's a very concerned, and I think potentially rightfully so concerned of you from physicians and patients on the trust of AI. And so from your vantage point, I mean, you're deep in this stuff.
If anyone is using AI, it's going to be you in clinical trials. We're seriously, where are we on this hype cycle? What are you actually using? Has it actually helped you?
You know, give us some... Give us some... the real world here. Yeah, definitely.
So I'll tell you, you know, we have used AI to look at, you know, eligibility criteria for clinical trials. And it was in Amros. It was another AI tool we used, which was nice in that sense that it was very protected and local. It wasn't on a cloud server.
But the challenge we had was that there was no communication with our EHR. So if you find lesion sizes that actually qualify for the trial, but those patients have a glaucoma tube, Rayhan, you may know this well, or they're fake or pseudo fake, or they're different criteria, or they're a retract to me done, those patients are excluded. So I think AI needs to be more than just reading images. And I know Ben said Amros can talk with EHR and imaging together.
I think that's really important for us, because I don't want to be looking at those pictures and looking at their records and then saying, oh, wow, this was a screen for... I actually recognized the patients just by their names because I knew all of them, like none of them are eligible. So here we are. We did all this work.
We got the system. We spent all this time and energy, the rent through the database, and everybody that was found from that was not eligible. So I think the key is that there's a lot of hype about AI, Rayhan. And at my meeting, Clinical Trials at the Summit, where we had the luxury of having both Ben and Ram on the podium.
And I asked the same question. I said, there are so many companies. There's so much noise. Let's filter the noise and cut the chase and cut the BS and get to the reality.
And I think the reality is AI is going to really help us, but it has to be done correctly, right? There's a lot of security issues. There's HIPAA compliance. There's all these privacy things that we have to do.
But I think now we're at a level, Rayhan, that we're actually almost there. So imagine, and I haven't used Amros. I have not used Tilda yet, but I'm really excited to implement it because the staff, I think, issue continues, for example, I have four or five staff, they are interviewing at med school. And most of them are assistants, but some of their coordinators do and have to keep replacing them.
And that's why I'm going to call Ram right away to, hey, help me because, you know, the biggest success of a site is in a staff. That's what Ram said, and that's totally true. You can have amazing doctors. You can have all the trials.
But if you don't have good staff, you're going to have protocol deviation. You're going to have issues with enrollment. And the data you're going to have is not going to be reliable. Now, that being said, you know, we have to implement slowly and realize where the weak links are for AI, right?
AI is not a be all for everything. So for me, it's going to be more like, okay, I did the first version of just imaging AI. I think the next version with Amros and Ben will be helpful where they're going to data mine. I have about five GA trials coming.
And I was going to call Ben anyways, because I want them to mine my data so that I have all those patients that are eligible that I can call about the interest in trials. And then, you know, getting AI teammates to replace some of the staff that are going. So I think we are evolving, Rehan, like we have not implemented fully these technologies for me to say that this was amazing. But I'm really excited.
And I think they're just going to get better with time. Yeah, I agree. And I'm an optimist here. I think that's definitely the case.
It's just a matter of timing. You know, I think we overestimate what's going to happen at two years and we underestimate the impact over a longer time period, like 10 years. And so, but then speak to that. Because I think for me as certainly, I'm not Dr.
Anani's level of clinical trial specialist, but recruitment seems like we need a special, almost a special conversation because that is such an obvious use case for the use of data at scale. I mean, as our show is saying, mining patients in advance of your EHR. So it seems pretty straightforward and eerie. But what are the pain points that you're facing at Amara?
So what is, you know, what keeps you up at night? How is this really going to work in practice? To me, it seems like, oh yeah, just mind the data, look at lesion size or whatever dry ice that is called coma studies. It seems like almost like a filtering on an Excel sheet, but this is like a very low level, you know, unintelligible way to look at it.
So Ben, like inform me, what am I not seeing here? No, it's to us. So we've been around for seven years and we've completed 36 studies last 18 months. So we learn a lot, right?
What not to do? What is, what to do? And we're very excited that Dr. Kanani is also joining our network, obviously.
The most important part wasn't the AI for us because that's part of it, the deep tech, we understand very well. What we knew that 80% of the optimology data, general is a healthcare data as well, that it's not either accessible, it's not structured, it's not available or misused, if you will. So that going clinic by clinic practice by practice, putting the sleeves up and integrating these five different data points, imaging is one, but not just imaging, right? So there's OCTs, FA, FAFs, and then having these image analysis, that's great.
But then we still have to go to the connecting the dots with the patient and per visit, which also requires medical history of the patient, insurance, and the provider nodes. So that merging, that structuring was the hardest part, and we believe it has been our part. So AI, it's, to us, it comes next, because once you have the data, you can make sense with it. It's just how do we get to that point?
Was the talking about the white stuff, that's been also our challenging, and I couldn't sleep for years, I would say for this one. But outcome is once it's done, is automatically doing this now on behalf of the coordinators or the administrators. You mentioned about the, you have been pouring out the staff, and what we do is we eliminate the time for coordinators and administrators to do the manual checking, if you will, for these charts. And when we give a list, and it comes, let's say, every Monday morning at 9 a.m., and they have the full list of who, which patients are coming next that week, and which ones they qualify for, which studies, let's say, right?
So we're taking that, we're saving time for the coordinators so that they can directly engage with the patient, educate them, and work on the randomization, if you will, a process. So that full automation, not requiring any coordinators to be involved during the process has been tough to build. But without that, it's not working very well, right? So then you're getting, again, is somebody has to do this, somebody has to do that.
Now it's very clean, the list comes in, they act on it, everything is good to go. What else I will say one more thing regarding the people, we believe human in the loop, not just the AI. And so each of our results, even though our machine platform provides the information, we still double check it with the humans, right? And without having that confirmatory system, we don't believe that's the right way.
Or the better way is both the technology and humans coming together and finding that solution, finding the results and answers that the clinics are looking for. And maybe Ron jump in here too, but how do you scale this even more? Dr. Nani, you have your universal patient, I'm sure you do outreach, I have very sophisticated ways to have outreach.
But the vast majority of patients who are seen for eye issues in this country are actually not seen by ophthalmologists like you are me, that they're seen by ophthalmologists in optical retail settings. Is there a future where we can use the tools from Tilda, Amros and others, to sort of get them from somewhere else to see you who's actually running the trial in a streamlined way? So I should speak to that, is that a future you're excited about? Or like where, what are the bottlenecks right now?
What's not working right now? Because I could see, you know, look, if you're a patient who has geographic atro and you're being seen by your opt-of-contrast, and how do you get that patient to you? Yeah, or whoever who's running that kind of trial? Yeah, that's a really good question.
It actually came in a conversation yesterday with a company that is looking at diabetic retinopathy, right? So a lot of diabetic retinopathy patients are sitting with optometrist or general ophthalmologists and because they have good vision and they don't have macular edema, they're not coming to us. So the question was, how can we use AI to pre-screen the diabetic retinopathy patients in optications, ophthalmologists, optometry practices without issues of privacy and all these other things that go into using an AI model, right? And that's a big bottleneck, right?
Like we are a research site, so we are set up to accommodate new things, right? If you are a busy optometrist or an optical shop, refracting patients, selling, you know, glasses and contacts, and then yeah, patients do get optos, images done, how do we get all those patients identified using AI and then referred to us? So I think that is gonna be the next way, Rayon. I don't think personally we are ready for it yet because we're not ready yet.
We are implementing in the retina sites, right? And then the next step will be, how do we convince the stakeholders that don't have us taking research to let companies mind the data? And now is there, of course, any work adjustment fees and things like that? And what kind of consents do we need, right?
I mean, what is the requirement regulatory-wise? If I'm looking at a database, somebody's data, and somebody's AI, even though they are de-identified and we are considering them for trial, do we need their consent before? We do that because anything related to research when you are being considered for research needs a consent. So there's a lot of regulatory and other things that as a field, we have to work together with the regulators to find out the right way to do it.
But I think right now we're doing manual stuff, as you said, referral dinners, education, hey, if you have any DR patients sent over a wave, you have great trials, you have GA patients sent over a wave. So it is pretty manual. It's kinda surprising that this day and age of AI, we are still doing those things. But as Ben said, humans are not going anywhere.
All of these are gonna be where humans are involved. So I think that fear that people have, that AI is gonna take the job away. I wanna close the door on the AI. I feel like open the door on the AI so we can become more efficient and more productive at the same time, don't lose our job.
So yeah, that's the second stage I think, where you ask her with the question, Rayan. Well, maybe I'll add something and then I wanna run maybe we can add also. Piggy Lekomar was saying, it's a referral system is a tough one. We started to change that with our Amaro's research network.
As I mentioned, the power of our expansion is any of the optomethological clinics, right? So we do work with optometmologists, optometrists, as well, this could be based on let's say in the where the retina specialist or go a convers specialist based on the radius as we have potentially referring clinics that we work with, we piloted that and it's slowly working out. It's gonna, as Arshad said, it's gonna take some time. I think the challenge is not identifying those patients at this from our end.
It's the how is the business model is gonna work? How is that consent? There are rules, regulations that needs to be adjusted for that as well, it's not just identification. But we've seen this in the multi-specialty sites that we partner with, some of them are private equity firm based, that they do own opticians, optometologists and specialists inside we've been doing it successfully.
So that we have some idea, but when you take out from the independent ones, then the different rules and regulations, but it has to be fixed. I think you will get there. Yeah, I'm sure it's imagine, you know, a young kid or maybe a 30 year old with bony speculas in the periphery who's losing, has poor night vision and it's seen by an optometrist. Maybe they don't have insurance and it's just sort of and tough kind of using these patients.
I mean, like, you know, they had big this and they're just not, you trained in North Texas, right? Where I'm at here in Dallas. And they want care. There is probably a clinical trial out there that would provide care for them during some and important time period and potentially help with their symptoms.
It seems like there's a lot of potential upside here. I don't know. Ben, is that, and it just seems like you got it. Yeah, that's what we're seeing.
It's just the challenges convincing them to go or to do, once it's referred to follow up. Like we've done my previous experience, we work with federally qualified health centers and we were screening patients for diabetic retinopathy or other eye challenges. And we were identifying these patients both for the connecting to the specific therapeutics or the clinical trial. It's just that explaining, convincing those patients to refer and follow up, you know, they're in maybe low income and transportation is maybe a challenge.
They have kids in the schools or multiple, you know, it's been working true jobs. So it's been a tough to move the needle in that regards. I think it needs to be some incentivized system will be very nice. I don't know what dancer for that, but it's not, to us, it's not an identification challenge.
It's what happens after that. I don't know. I don't know. Yeah, you know, we're not directly in the recruitment problem.
So maybe I can't directly answer that particular area. But I think I will, I have some really interesting learnings on the previous discussion, which started all of this, right? Which is how fast are we adopting it? Where is the hype cycle?
Where are we in the hype cycle? So where we stand from a vantage point, I think I've learned a few things. First is AI is really exciting. Everybody feels that excitement.
There's a lot of like formal going on. You want to be on the train, not left behind. So we get it, right? We see this all the time.
We speak to so many sponsors, CROs, sites. And what I have realized is, there's a couple of things which most people are, you know, either saying without saying it or it's just there, like if you're implicitly or explicitly there. So first is there's a trust issue. Most people say, well, I know AI, I use charge GPT, but this thing makes up stuff all the time.
Like there's hallucinations left, right and center. So why would I pay you to basically supervise my, the day or the whatever the outcome by myself, right? Like there is this like clear like mismatched sort of like situation which I think if you don't think about the product in a proper way, you end up in that zone. And I think I love what Ben said, supervision is so core to the full idea of how do you, how do you build successful updates and AI products, right?
So I think what I have noticed is, you know, we've definitely hit that S curve in the last couple of quarters. We're in a pretty large growth curve right now. If you haven't seen us yet, you will see us because we work with about 50, 60 plus sponsors, CROs right now. They're reaching out with thousands of sites with our tools right now with our AI teammates.
So it's going to be very prolific in 2026, you know, almost every site will interact with an AI which is built by Tilda given the proliferation, especially on Kumoji, we have a quite a bit of momentum right now. So how do we get there, right? I think for me, it's been a couple of winning solutions. First is I go in and we say, we're not a build process.
We are a buy process. So you don't, we're not coming in with a shiny toy, which then will spend the next two years perfecting for your use case. We're not a consulting shop, which is selling you a half page solution way. This is the productized AI model, which is that the state of the art benchmark in this industry.
And we've done a lot of publications. We've worked with our existing customers who love our products and are now talking in conferences and publishing with us writing papers, posters, that sort of thing. So I think that trust needs to be earned. Like there's no, there's no shortcut to that.
You need to go out into the market and you have to start somewhere. But you have to start with a product, which actually is something which can stand behind and say, this actually does work. So I think that's taken us some time because, you know, that's not an easy thing, where you just kind of like, you can't just launch and then get all that. It'll take sort of like the early innovators to come in, early adopters to come in right after and so on and so forth.
And we're seeing that momentum, which is really like worked out from a referential perspective. The second I would say is ROI. You know, all said and done. If you look at a lot of the, in the hype cycle, a lot of the companies or people who are talking about AI, these are not tools which are really at a place where the ROI makes sense for most of the solutions right now.
Like you're being asked to spend millions of dollars to implement solutions right now. And that's not a budget which most people care about. Why would you do that unless you actually have a fully fledged product and an ROI story with a never a shock to your plan? You will get a handful of them.
Like, you know, the top, the biggest sponsors out there are probably working on these like Palantir style contracts where, you know, you're coming in, bringing like 50 people, 50s and building whatever you are built. But that's not gonna lead to like that option we're talking about. Like to bring in some of these incentive models with Dr. Kanani and Ben or Talking Board, very widely distributed AI everywhere.
It has to be perfect, it has to be good, it has to be very quick to ROI. So that's kind of where we start. We just said, okay, like, you know, just try us out. You can pilot four to six weeks.
You'll know this works. This is the real thing. It's the real deal. It's got a austerity art performance.
And if you like it, you go live right after, right? And what we've noticed is we always compare our pilots against a human controller because, and I'm talking most of them as sponsors since here, I was prospective here, but in the workforce, which we manage like starting a management or regulatory or data, we generally go with like a model where our pilots are on active studies, which they're already working on. And what we say is because these are active studies, you already have a human controller. Whatever data you're generating from your processes, we can compare ourselves against those in metrics, right?
So how a strategy in terms of like figuring out a look where you can come in and you can sort of build a comparator study very quickly within like four to six weeks. And if you can show that clear demarcation within that timeframe, where your quality is and so coins better than human control arm, your time to resolution is like 100 minutes versus like days, so there's that advantage with AI. You have an excellent supervision model, which again, the way I described it is, don't buy the cruise control solution. That's like you supervising it and you don't know when the car is gonna just like do whatever it wants, right?
You want the other way, you want the way model self-driving taxi solution where you get in the car, you're trusted, I take it all the time with my kids, like I feel really safe and I blame all, it's fully autonomous, right? There's no driver, but I know for a fact there's somebody somewhere monitoring exactly what's going on with this car all the time. They don't need to be in the driver's seat to do that. There's supervision models, you can build very sophisticated multi-layer supervision models right now along with human supervision, big into it to be able to like guarantee that quality.
And that's kind of like all the type of stuff we've done to kind of get a model, right? So I think the necknut of what I'm trying to say is I do think the momentum is absolutely there. We feel it, we're certain that AI teammates are gonna be everywhere like into next year. Just, I'm just talking more from my perspective but I'm sure there's others as well for seeing this.
And I mean, it's really exciting. I think we're gonna be like creating a new type of incentive models, new type of operating models as well. It's gonna be a race because you don't wanna be left behind and there's just, we're gonna just continue to bubble up into like something much bigger than where we are today. And I think when you see the sort of like earlier operas it's very clear to me that they're already like moving very quickly away from whatever their plans are from like a non-AI approach to like an AI approach, right?
So I think like in general, I don't think we are actually underestimating our, I don't think we're overestimating the two-year cycle. I actually think we're underestimating it. That's where I sit. Hey, a little bit of my kinesome.
I love it. Let's say. As we wrap up, I had not used Waymo. I love my Tesla self-driving.
I trust it because I'm behind it. I'm not, I've been in a Waymo. Maybe I'm not courageous enough. Next time I'm in the Bay Area, I'll try to wrap up.
Maybe we'll just go around here. And this could be more, I think, generally clinical trials. If you want to talk about how AI can help go for it. But I was like to end by saying, by asking, all right, what keeps you up at night, Dr.
Penani, on this, on a clinical trial topic, on AI in general. I mean, it is a tough landscape out there. I mean, just from a general economic time, there's a lot of anxiety. So what keeps you up at night?
And what gives you hope? In the context of AI and clinical trials, and sort of what we've been talking about tonight. Yeah, I think looking at AI, I think, what keeps me up at night is to make sure that we're doing it the right way. We are making sure it's secure.
We are making sure that it's identifying the right patients and doing its job, right? So I'm not worried about AI taking over my job because I don't think it will. But I mean, it won't be as good looking at it. I mean, in Belgium, I mean, let's just be honest.
Him can make a teammate that's way better with me. So maybe already, the other will have them. But the other thing is that we have to, you know, as Ron was saying, the hype starts and everybody wants to ride that hype and has FOMO if you are not part of it. But at the same time, what keeps me up at night is we have to do it properly.
And we have to do it the right way. I think that's where I worry about that we have all these systems that wanna, you know, partner with us, which one is gonna be the best one in terms of security, in terms of, you know, everything else. So it's more like not the medical side I worry about. Actually all the ethics and regulation that they keep me up at night.
What gives me hope? It's amazing. I mean, you're brilliant people like Rehan, you, Ben, Ron, we're all working together here to move the space forward. It is gonna happen.
It's just a matter of time, but we just have to do it properly and not rush into it. For me, I'll have a little different perspective, not the AI, but more on the human side, the coordinators and the administrators or the clinic side, because we believe, at least for the recruitment, AI is here. We are very comfortable with, we've seen the results. The challenge that we have, or what keeps me at night, is once we provide this eligibility list to the practices, that workflow has to be adjusted or maybe changed based on the new AI way, let's just say.
Example will be is the previously before Amaros, the manual search chart review was let's say X amount, three, four, five patients, let's, I'm just giving you an example. We are providing, let's say 27, 30, 17. So we're tripling quadrupone and decreasing the screen failure rate. So the delta is huge, which is great, but the challenge is if the practices can act quickly on this, the patient may progress and they may not be qualified anymore.
So our challenge is how do we work with the practices, which we have great response, of course, some of, sometimes, of course, some people are a little different, takes, there's a traditional way and takes their time, but that workflow, I believe, has to be adjusted as well, not just the AI, it's what comes after that, what are the humans that are doing with that AI? And to me, it's the, right now for us, the most education is that towards that direction and explaining that this is not a scary thing. Let's all work together. This is a new way, we're on their side, we're on your side and let's figure this thing out, is why am I?
And what's exciting is it's happening. The results are there, it's slow, but it's coming. As I said, I think the 20, 26 is going to be very, very good here. I'll finish it off for you.
Yeah, I think what's keeping me up lately is I fully understand the hype cycle and it's kind of like, you know, you have a syndrome where you feel you have the best product, but then there's all this type around you and you feel the need to differentiate and actually like stick out into the market saying, okay, I am the guy, I've actually got it, right? How do we show ourselves to be that way? And so that sports, I would say the big challenge because you don't want your customers or potential customers to be burnt and then just get this like, you know, the crash on the hype. I mean, you want the market to evolve, you want it to evolve the right way so that people actually like are like getting a great experience out of all this great technology and innovation we're making.
So obviously from our perspective, we don't want to put our name on any product which is immature. So we don't, you know, that's part of the solution there. But that doesn't happen very often in a hype cycle. The hype cycle's been everybody's saying everything's out there, right?
It's ready. So that's part of the challenge for us. I think the core angle here is that there is so much incentive right now to drive towards AI adoption among, you know, the C-suite, among investors. There's clear alignment of economics and motive, right?
Like this is a very unique moment of why this adoption is happening so quickly on the AI side. And so I think at some point what will happen is you want your success or your customer success to be promoted. You want your customers to go and talk about it and they have a champion the story and they need to go out there and talk in conferences, there's published and listened to that. And I think if more and more people and more and more companies like ourselves really promoted that kind of vision, I think we can really solve for that high problem because that's not easy to replicate unless you have something.
So, you know, I think I have thought because we seem to be doing parts of it. And of course, some of the great other AI players are also doing that. And I really like, you know, follow that and making that happen is probably one of the joys of like what I do right now. Seeing our customers be successful and them getting recognized.
Well, fantastic. Well, what keeps me up at night is my my four year old and my one year old. And my hope is out there to grow up and just sleep through the night one day. Hopefully AI, I'm people with AI robot.
Are you guys working on something like that? Like take care of kids? That's what I that's an art. But I think you have to do certain things by yourself on this one.
That's the joy of life and is going to wait. Well, thank you. Thank you all for and thank you, Dr. Nani Ben, Ram for joining us for a special episode of the AI innovators.
I really appreciate your time. And to all this, please check us out at ai and I care.com for this fascinating and really I learned a ton and many of our other videos there as well. Thank you again, gentlemen. And looking forward to seeing you around at the conferences coming up.
Take care. And thank you. Bye, guys. Thank you very much.