All right. Well, welcome everyone to another episode of AI in Eye Care Innovators. I am very pleased to have two good friends and experts in AI and clinical trials, so kind of like my two love interests. This is a topic that's very near and dear to my heart.
We are very welcome to have again back, Ram Yalamanchili. He is the founder and CEO at Tilda, which is a biotechnology CRO firm based in the Bay Area, and as well, happy to have a good friend of mine, Dr. Andrew Pucker, who's the chief development officer at Mintra. So just by way of introduction, Dr.
Pucker, I would love for you to just talk us through a little bit about your background and how you got into your role at Mintra and what Mintra does. Sure. So thanks for having me, and it's good to cross paths again. So I'm Andrew Pucker, I'm an optometrist, as Rehan said.
I went to Ohio State, and I got a master's degree in tear film chemistry while getting my optometry degree, and I found that I liked science so much that I stayed on, got a PhD, and that went well. I got a K08 award, which was a transitional scientist grant that helped get me a faculty position at the University of Alabama at Birmingham. And I had my own lab there for about seven years. I was in clinic part-time, doing research part-time.
While in clinic, I was running the myopia control clinic and working in the cornea and contact lens practice. But I was also doing dry eye myopia research and kind of a little everything in between there. And it was split between NIH research and industry research. And as I went along, I learned that I liked the industry questions more than the NIH sort of stuff.
So I made the jump to a company called Lexitas, that is an ophthalmology-only CRO where I actually met Rehan there. But I went on to lead the anterior segment disease portfolio at that company, and I did a lot of protocol writing and everything related to that. And then I saw this opportunity at Mintra Health where they were building an automation-forward CRO. So what they were doing there is, again, doing an ophthalmology-only sort of clinical trial space thing, but also working in AI technology, all those good things, to try to disrupt where we're going with clinical trials.
Like we've been doing things manually for years, just like reviewing data by hand. All that is really monotonous and probably pretty boring for the staff, right? So we are trying to use technology. For example, we're partnering with Tilda, which is Ram's company, to make things more efficient, better quality, and overall, just more fun.
Fantastic. And Ram, when I introduced Tilda, I think I made the same mistake I did the last time I had you on, where I limited the scope of your company. So I apologize for that. So rather than hear from me where I'm just getting things wrong left and right, my wife tells me that as well, why don't you tell us what Tilda is and what Tilda does?
Yeah, absolutely. And thanks for having me here, Rehan. Nice to see you again. So I'm the founder and CEO of Tilda Research.
What we do is we build AI teammates, as we call them, for the clinical research space. This is, in other words, the way to think about it is, clinical operations is heavily manual in the way we perform work today. And there's quite a amount of tremendous AI capabilities which have emerged in the last several years, which we're then using to automate or bring quite an order of magnitude productivity gain in clinical operations. So when Andrew's talking about building a CRO on top of our solutions, we are also not a CRO.
We build the AI teammates, and we are mainly just on the building frontier models for clinical research and the harnesses and things like that, so you can have clinical operation leaders use our tooling to be able to do this much faster, better, cheaper. So, that's kind of where we are partnered and how we're helping Mintra and Andrew's team out there. So dive deeper a little bit. I love the term AI teammates.
I think that's a great phrase. I'm going to steal it from you actually, because I loved it so much. Maybe our next column, I'll put that in. So tell-- dig deeper.
What is an AI teammate in regards to a clinical trial? How does that work? Yeah. I think, just definitionally, if we just step back and sort of look at what exactly is an AI teammate and how do you define it, I think the best way to imagine that is to have a AI which is acting on your behalf in certain areas where you would like to delegate work to.
And we do really mean delegation of work, just like you would delegate work to one of your colleagues or something like that. It's a very similar type of a concept in terms of how we build our products, right? So, in other words, it's different from, say like I have a software where I go in and I do the work, and I basically record what I've done into the software. That is the traditional approach of building software, where I kind of look at them as sophisticated filing cabinets.
So you're still doing the work, and you have a team of people doing the work. And once everything is done, you're essentially cataloging all your work into that system. So that's the typical paradigm of what's happened with software prior to AI. I think the place where we're going, and not just in clinical operations, in general, in many of these services industries and even personal life, is we're all going to get used to this agentic era, which is rapidly coming upon us, quite a few consumer products which are coming out right now, which I think all of us will get used to in a couple of months to a year or so.
So sort of think about it the same way, right? There's many opportunities in the day-to-day of clinical operations where you can say, "Hey, this is the work which I don't need to get involved. I would rather delegate it to somebody," and that somebody is just an AI teammate, which understands the context, which is able to go into the systems you already work with. It understands how to use those systems.
It's broadly really capable and well-trained on the concept of clinical operations. So, like for us, that's where we come in, right? What we are saying is, let's take an AI model, which is a generalist or a foundation model. It's a world model, right?
So it understands the world in general, like general knowledge. But then let's post-train it to be even better in a certain domain. So that would be, in our case, a clinical operations domain. So that would be in various forms.
We have four different pillars in which we've trained these AIs, and we have four different AI teammate products basically. So you can sort of think of it as like one category is maintaining your regulatory documentation. So the entire TMF management, audit, completeness, inspection readiness, just give the AI teammate a bunch of work, and it's looking through the documents, filing into the right places, making connections, building a knowledge graph out of it, figuring out is there anything wrong with this picture. Like maybe we've done this, like 3,000 document uploads, but then something is wrong with these like 30 documents, or some other state is not quite adding up based on what I've just read through these 3,000 documents, right?
So this sort of like very high-level, complex hard checks are really easy from an AI perspective to manage at scale. So we've got-- that's one of the pillars. Then we have something in the site startup and maintenance. So, as Andrew can probably talk about, we spend tremendous amount of time essentially working in research to bring sites onboard to work on your trial.
To the point where, from the day of contact, like feasibility all the way to first patient in, can itself be like a year, a year and a half in certain cases. There's many complex scenarios going on there. There's budget negotiations, feasibility through the various forms. But you can by and large really reduce the amount of touch points, which a person has to do, by automating or at least giving this task from AI teammate.
And what we've done is basically gone into nitty-gritty detail, like really trained them into these kind of workflows across millions of data points which we have. And then the other two pillars are data management, so quality checking, working on the data which is coming back, looking at queries, creating queries, managing queries. And the very last one, which I think is actually a very interesting and big pain point, but not necessarily on the research site, is payments. So, research is great.
I think we all need to encourage more people to participate. But one of the big pain points is, it's just a very tough place to really understand how you're processing payments and how you're getting paid. And there's a tremendous amount of overhead which comes in, in every organization performing clinical operations to figure out what is it that we owe? How much do we have to pay?
Is everything we've done correct? And even forecasting, like what's my outlay into the next three months to six months based on what's happened at all my sites across all my patients? Because not all costs are predictable. There's many incidental costs generally.
So this is a very complex problem, and what we've done is basically gone out and worked on these four pillars, and built basically AI teammates which can then delegate some of that work to that. One of the things I'm most excited about is this startup agent that Ram is talking about, and I think it's just going to make us look better at Mintra. It can kind of sift through what we already have on file, and then we don't look bad by asking the same questions all the time. Maybe we have an active site, we've used them several times.
We already have their practice location on file. The system could potentially be written so that we already have that at cross-references it, and then we just ask for stuff we don't have for that specific study. So, and could you give me an example of a dry eye study? How does it actually work?
Just to put some concrete teeth on it. Yeah. So it's not uncommon for CROs to have multiple studies in the same indication. If we use dry eye as an example, there's maybe five studies that you're running related to dry eye.
They have very similar requirements. So like, what type of patients do you need? What type of study staff do you need to have to complete the trial? And if you're trying to gather documents for each of these trials during start up, the first time, you have to ask for everything.
And then the next time, well, if we just started a site up for a different dry eye trial like two months ago, we can kind of cross-reference that and just whittle it down to a much smaller number of documents that we need. Like we only need one CV potentially for the year from a doctor. We don't need to ask for that every time. And then it kind of helps us to not be redundant.
We do stuff and maybe we forgot we did that, so it'll help us organize ourselves and file things at the same time. So I think it'll be less burden for the CRO staff, less burden for the site, and then overall, we can focus on things that we really care about, that like quality is important, but I'd rather be looking at data and exciting things like that. Yeah. And maybe you're about to talk about getting the right patients for the study, and that's probably a big issue, right?
So site startup is great and all and making sure that all the documents are all there and filing is appropriately done. But how about subject selection, and making sure you're putting the trial from a sponsor side, you're putting the trial in the best position possible to succeed? This is something we're still working on, and Ram can maybe fill us in a little bit where they're at with that. But one thing we want to do eventually is have an AI teammate that can look at your EDC, your data capture system, and help us sort, does this patient really qualify for this study?
And in dry eye in particular, there's often a number of inclusion criteria that are pretty complex and easy to miss because they're on different pages in the EDC system, and as the doctor, you're only seeing one page at a time, but you could potentially make a teammate that could cross-reference all of that data, look at it, and maybe the data are even looking like they might fit the mold, but the AI could be smart enough to ferret out, "Hey, maybe this isn't exactly what we want." And I'll kind of turn it over to Ram here since they're building that now, but that is something we're really excited about. Yeah. I think, trials are complex, right?
There's many areas I think which can and need to be optimized to get the end result. So I kind of look at it as like a, if you had a complex multi-step process, optimizing one step in this multi-step process is not going to give you overall productivity. It's just the nature of it, right? If you have more dependencies than the...
That's not how it works. You either start deleting certain processes or you have to overall reduce the burden across the process. There's only two ways you can get productivity out of it. That being said, patient identification is a huge problem, for sure.
I think the way we thought about this problem is you have two ways to sort of manage trials, and trial bandwidth, because for us, at a macro level, what we really think is we're going to have an explosion of new drug candidates coming to the market. There's many reasons why that's happening. There's macro reasons on the funding and things like that, and there's also AI-driven solutions and other things which are probably going to give us more targets to go for in terms of trials. I think the constant, however, if you look at the last 10 years, is the bandwidth which is available right now to actually perform the trials.
There's only a certain amount of infrastructure which exists, certain number of investigators who exist in the space. And so, we have not seen a material increase in the bandwidth over the last 10 years. If you just look at it from a pure infrastructure perspective. So in other words, if I look at it as like the railroads or a number of airports, like how much capacity, how many passengers can we actually move from point A to point B, in these infrastructure equivalents.
And there's a certain capacity and you drive capacity up, and there's a certain growth curve around that. I think clinical trials is one of those industries where we have not seen a large increase in the bandwidth. So we have to solve that, right? And the way we think about it is a couple of ways.
One is you reduce the burden on what you're already doing today. So this can be all the way from the most mundane things. You don't need to send the same document over and over again to the same CRO. Just as an example, we've recently read out a use case for one of our large CRO customers, and they're managing 26 different studies across 1,300 sites.
And they've basically delegated our AI teammates to go do the startup process for these thousands plus sites, right? It's really amazing. They've got 15 different countries, eight different languages across these countries. So you can see this AI essentially talking in languages from English all the way to Portuguese and Brazil and Spanish and German and Hebrew.
And this is quite interesting to kind of see how work is now changing from, like you can do a lot more with these AI teammates. And the net effect of this is as you're doing more startups across this footprint of sites, the next startup at the same site is up to like 50% lower in burden because you just went to that site and asked them for all the information you need. It could be even feasibility. Do you have the right set of patients for this particular study?
They answer something. You may say, "Okay, well, this may not be a great fit for you," but the next time, I know the next four sites, I don't need to know feasibility, right? Sorry, the next four studies. Yeah.
Because I already have the answers. Same thing with startup documentation. You gave me all the investigators who are present in your site, their CVs, their trainings for a study which you're already active on. And there's a tremendous amount of overlap in this sort of activity.
Today, we just do this over and over again. It's just such a low-hanging fruit, which we can kind of very quickly address. And so the net effect is we're starting to see startup times really collapse at this point in these kind of examples. We're starting to see months of time collapsing.
And what used to be previously a linear curve of startup, so in other words, if you were to start up a Phase III, 100 sites, you're going to go in phases. You're going to go first 10 sites, 20, 30, 40, 50, right? And largely, it's dependent on how many team members you have surging into that requirement. But today, if you have an AI teammate essentially being delegated, you can say, "Okay, instead of 10, go ahead and do 50."
And if everything's good, then we'll do another 50, right? Or you can do 10 and then 20 and then the remaining. So there's no sort of reason to go with the traditional model of how you plan bandwidth and people resourcing. And so what's exciting for us is this is happening on the sponsor CROs with our solutions, and we expect the same to happen on the sites as well.
Like sites will also have agentic solutions, will have AI teammates of their own to manage these type of processes. And so, we are entering a place where agent-to-agent interactions are going to become more and more common. The industry is just... I would say most other industries have already moved on there, like I don't know if you've had a chance to recently call customer service at United or something, and it's all automated, including the voice agents are automated, right?
They're all AI. And they're quite good. They're actually able to solve these problems very quickly and get you to the end goal of... Unlike like six months ago where you just want the real person on the phone.
I know. Six months ago. I noticed that, before I would see like zero, zero, zero, zero, as much as possible trying to get to the real person. Yeah.
And now, it is interesting. They are getting a little bit more autonomous where they're actually making-- They can make a, like do a quick return. I don't know what their rules are, but it is, I think as a society, we're becoming more accustomed and amenable to these types of interactions. And as you mentioned, like an agent-to-agent.
So are you imagining- On the site, and we talked about sponsor side, we talked about the CRO side, and you as a sort of intelligence platform aiding the CRO at Tilda. What about on the site level? I was a private practitioner for many years, still am a private practitioner, and being a site is hard. Oftentimes employees are doing multiple tasks.
So how do you support the actual sites? Yeah, I think that's a really good question, and I really like what you just said about being accustomed. I think that's the key here. We're just entering the phase of being accustomed to working with AI teammates, right?
We've been in the industry, I've been doing this for many years now, so I'm seeing the evolution happen. I would say end of 2026 is when a majority of people are starting to really understand, "Okay, I've actually just spoken to an AI, and it did the work." Right? "It actually got the result of what I wanted."
I think we're seeing this in customer service is probably the biggest touchpoint which we're seeing. But there will be other industries, and it's going to really rapidly grow because a lot of the phone manufacturers like Apple and a bunch of other guys are also bringing these things very tightly coupled into the operating system in the next couple releases. So we'll just start to see this really big shift in terms of being accustomed to then AI being your delegating partner for doing certain things. And anyway, so I think my point here is that everyone will eventually get there.
I think it's happening really rapidly. And to your point of how do sites manage and how do sites adopt, I think, see, my perspective here is sites are really burdened in every way possible. Everybody talks about it, right? The coordinator is the most burdened individual in the ecosystem of research, and we can't find enough of them.
We can't retain enough great people to be in the industry of coordination. I think the churn is like some 60% something I've read somewhere. It's a very high churn, high burn type of industry with very little-- the metrics don't look great. So I do think these type of solutions are a good fit for an industry at a level.
Where the challenge which I've seen and which I think about is what does adoption and diffusion look like of these technologies into such organizations? Because one of the things about change is you have to invest in change. It's not easy to sort of just adopt something and move towards it. You have to have the resources.
You have to have the top leadership. A lot of things have to come together where you are deliberately moving towards this new world of using AI teammates. And I think what we're seeing in general in the industry is the people who are best positioned for this is either individuals, so like people like you and me, who are basically saying, "I'm just going to use ChatGPT for a bunch of stuff," or whatever it might be, right? So we are adopting personal agents and things like that as they come in, and we're seeing that adoption curves rapidly happening.
And so there's quite a few companies which are getting built very successfully right now. So individuals and I would say mid to large companies. So this is where the current adoption and diffusion is happening because primarily it's the investment either on time or resources or both. So this is where it is.
I think the middle ground is where small businesses and sub-20 people, sub-30 people employee companies, how are they going to transform themselves? What is the investment cycle required for that? So these are the kind of questions I think we have not yet figured out as an industry, and I think it'll take some time. I think solutions have to be at a point where it's very easy to adopt with very little change management.
So I think it'll take some time, but I expect this to happen over the next couple of years, right? So it'll be in a form where it's not in the current form, but I think most of the solutions need to get there from what I've seen. So yeah, I think that's where we are. We think about how do we get there, what are the type of solutions and how do we position, how do we package the entire solution from a site perspective.
Currently, we don't really do too much with sites. But certainly thinking about it. I totally agree with Ram. There's always going to be early adopters, trendsetters, late adopters, but as things get better, those late adopters are going to just start getting into the fold because they realize it works.
And as AI works, it keeps building upon itself, right, Ram? You kind of teach it as it goes along, because I think it's only going to get better. Yeah. Ram, you mentioned sort of this capacity problem and clinical trials, as much as we want drug development to proceed, and that is a bottleneck.
And one thing, maybe this is specific to ophthalmology, but it seems like the number of sites, and maybe just going back to the site question because I've been thinking about this a lot, but how do we get more sites in the system? And Andrew, I know we've talked about this before, and in a previous life we had sort of discussed this, but it's a common-- many ophthalmologists and optometrists will come to me and say, "Hey, Rehan, I would love to be a site for a research study. I see a ton of patients. I'm doing this, that, and the other."
And I think this is an area because especially as we're talking about rare diseases, genetic diseases, we need to be better as an industry in being more heterogeneous in our patient selection. And if we're just doing the same dry eyes sites over and over and over again, and we all know that. Sometimes I wonder, can we do a better job, Andrew, of getting more sites? Because I think there's interest, and it's just like, why is it so hard?
It is challenging in that a lot of companies want people with a track record. So it's kind of like you got to get the experience to get the job, but you don't have the experience to get the job. It's a tricky thing. So what I like to do on our larger studies is have our established sites and try to bring in a couple of new sites to get them into the system because we need to keep building our site list.
People unfortunately are retiring, right? So there's good sites leaving. We need to keep new ones in the pipeline. Other things we could potentially do is do training sessions with different coordinators at sites.
They're the backbone of these projects. If you don't have a good coordinator, even if you're a key opinion leader in the field, you're not going to be able to succeed. So we need to have more coordinators that we already kind of talked about, and maybe the AI teammates will help with that. Maybe it'll be able to someday take a study coordinator who's not as experienced, but could also coach them along the way.
It could send them emails, "Hey, you got to do this today. Please do it." I think that might actually help get us over the hump a little bit, but yeah, it's tough. And one other thing I've seen with some sites is they don't see the economic benefit.
Yeah. So maybe a site might have a little bit of an interest, but they're like, "Oh, I'm not going to make money from that." And they actually can, but they just need to think about how they're going to incorporate that in their business. Yeah.
I just had on Dr. Sichther, who is a professor at Wilmer Eye Institute Johns Hopkins, and we're talking about surgical education, and I think we could translate some of those lessons of education to sites and coordinators. Ram, you had mentioned clinical site coordinators may be the most underappreciated and overworked people in the field, and if we could provide more education for them and augmentation, that would be great. One thing that always comes up in these conversations is if AI's not going to kill us all, before it does that, will it take our job, right?
And so I want to give you guys both a chance to talk about-- because you're using all the words. I'm using the efficiencies, augmentation, agents talking to agents, and I assume the answer is augmentation, not replacement. But speak a little bit more about giving these site coordinators, the sites, the CROs, Ram, for you, superpowers to really expand their skills. I think there is going to always have to be a human in the loop, specifically the sponsor, the CRO, the site is going to be responsible for the data in the end.
And I like to think of the AI teammates as a smart coworker that needs coaching, and it may not always be right. So you have to have a human check the data, and you can decide how frequently you need to check that data. But I think it's going to really streamline things, as Ram was talking about before. You don't have to have a human file hundreds, if not thousands, of documents directly.
They can just review them, which is way, way faster and more enjoyable and less likely to result in errors. But I'll let Ram chime in here. Yeah. Rehan, I have the fortune of actually looking how this is being played out at a certain scale in research, specifically at Rentronix, because we have fairly large customers who we work with.
I would say across the board, first of all, I'll just make it very simple. We have not seen anybody cut or reduce headcount. We've actually seen them hire more people into their-- across the board. Every single customer I've worked with, I pay close attention to this, so we've seen this happen across the board.
So then we ask why. This is the counterintuitive argument. If AI is augmenting or if it's doing a lot of the things which you are supposed to do, how come they're now adding headcount? What is the reason?
So I think the first thing is the bars are going to get set at a different place than where they are today. And by that what I mean is, as I was saying, some examples around site-related workflows or let's take the example of a TMF. One of the easiest ways for us to go and sell into this market of why you should not do TMF without augmentation, without our AI teammate, is let's just do an audit. Let's do a pre-PLA or pre-IND audit of a study you're about to go for.
And I think in 100% of the cases we've done this, we've done so many of these right now, we found severe issues. There are massive quality issues which are just un-- We just don't know because it's just we've hit the human limit of whatever we can to deliver quality at that scale. And this is usually with what the industry calls two eyes or four eyes processes. So there's two people, or at least three sets of people who might be going to checkup and check and check with periodic checks and quarterly checks and end of submission and pre-submission checks.
So I think what we are seeing is you just cannot use those bars anymore. The bars are going to set much higher. And so does that mean that now AI is doing everything? No.
AI is just helping you get to this high bar the industry is moving towards. And because everyone's going to have to now match up to that bar, their option of AI will happen to be that way. And the human in the loop still has to exist. You can't just have AI just do whatever it wants.
We're not at that place where we achieve AGI or the AIs are so smart that they're just allowed to do whatever the person is doing. So I think there is, in that world where you would spend 40 hours a week looking at a whole bunch of workflows and trying to do your best, maybe coordinating with 20 different sites, answering every single email, every single document which comes in, following up, making calls, this, that. Today, for the same amount of sites, you're still doing maybe an order of magnitude about the same amount of work, but your quality of delivery is way higher because you're now using your AI to delegate a whole bunch of stuff and you're managing it. Your skills have now transformed from you doing the work to managing somebody else's work, which the AI is essentially doing.
So I've just seen this transformation happen in front of us, in the last year, year and a half of this adoption curve. We've seen every single customer where they're saying, "We're just doing much better in quality. We're actually faster as well." So we're competitive in the industry.
So we're winning more as well. We're going out to customers. We're telling them we can deliver this kind of a work, which previously was benchmarking out here. And if you don't believe us, here are some case studies.
We can even audit it for you and show you what it looks like with bringing any of your current studies. So for us, it's become a very easy thing to establish. We can show this across communication traces with site startups. We can do it with data management, the quality of your data.
We can do it with your payments data. We can do it with your TMF data. It's so easy to sort of see where we have hit the limits. The bar is not very high at all given where AI is right now.
So that's it. It's like if you take any other industry, just because we had industrial automation, are we just all shedding on our ability to work in factories? No, we're just producing more. We're consuming more, we're producing more.
We have more complex things we are producing today. Every single thing you touch today is probably way more complex, millions of times more magnitude complex than what was in the 1700s. So we've just gone to a completely different field of expectation, and that's exactly what will happen with almost everything we do today. I don't know how exactly it will morph, but we're going to have better and bigger expectations in terms of what we want to.
I think, I'm trying to remember who said this, but there's a quote which sticks with me. It says, "There is no limit to human ambition." So unless we think there's no more, like we've reached the ambition curve, and that's it, then yes, we don't need to do anything more. But I feel like we'll just come up with newer things and more things to do, and more complex ways to do it.
Yeah. And adding to that point, I think this is a good place to segue because that was a very hopeful, optimistic point. What are your hopes for AI in clinical trials in the next, let's call it five years, Andrew. So what is one or two things you're just like, "I'm really optimistic that it could help in these certain pain points," and specifically maybe even Mintra is working on some of those?
I think one of the biggest things that we're going to get is better data quality in our data systems. Specifically, we have data being collected by multiple sites. They enter it in from maybe paper source or maybe they're using eSource, maybe they're a trend-setting site. But there's often errors in that, and we have to send clinical research associates to the sites to look at all of that data manually, make sure it matches up.
And we sometimes don't catch errors until maybe a month or two later. That's unfortunate, but it depends upon how frequently we're sending someone to the site. I'm thinking AI can catch those almost in real-time and make it so that, one, it's just cheaper because we're not having to have a CRA on site all the time, or maybe they just go less frequently. I think there'll still be a need for them.
But they go less frequently. We're going to be more ready when it comes to database lock. So there's less of a scramble at the end. You've been on studies.
I know you've gone through this where the last week of a study is just crazy. You're trying to put everything together so that you can lock the database and then run your top-line data. I think that's going to be less stressful overall. And then in the end, it'll be a higher quality, I think, because the data will be more likely correct.
So that's probably what I'm most excited about. I'm a data person, so that's what I'm really excited about. But there's, I'm sure, many other things we'll get from AI as we go forward. Yeah.
I'm very excited about that because ultimately higher data quality, you hope you'll have a cleaner signal to derive potential benefit for whatever therapy you're investigating. So yeah, I'm very much on board with that. I'm very hopeful as well. Ram, what about you?
What is the next five years? What is the one thing you're like, "Okay, we may not have it now at Tilda, but the next five years, I'm very hopeful that we're going to get this thing." Yeah. I think you touched on that, Rehan.
I think we're going to have more individuals participating in research, doing the research work. It's such a critical-- I just see this as a real challenge. We are obviously at a bandwidth problem here. We're obviously at a juncture where we cannot support any more than what we're already doing in best effort.
So reducing that activation function for the next new researcher coming onto the platform or into this industry, I think is going to happen in the next five years. We're just going to have solutions where they can perform a very high-quality amount of work with very little burden, and the amount of time it takes for them to get there. The trust factor will get built with the industry as long as they're using these tools. So they're going to see it as like, okay, you're new.
I understand you've never done this, but since your systems are so good, I think we are okay with essentially bringing you on because we have some trust in terms of what's happening and how we maybe can monitor this. So I think that's probably the most exciting thing about what can happen in the next few years. At least from my perspective, we have to solve for the bandwidth problem, and I think this is one way to do it. Yeah.
I guess you heard it here first. So better data and more people working in data and in research in general. Ram and Andrew, thank you so much for joining the "AI in Eye Care Innovators" podcast, talking about clinical trials and AI. Super exciting topic.
We're going to hold you to those semi-predictions, and maybe we'll come back next year and we'll do this again, see how things are going. Thank you again for joining us, and we look forward to talking soon. Thank you. Thank you, Ram.