E174 | How AI is Transforming Regulatory Submission and Medical Writing in Biopharma

We keep hearing about wave submissions and the thinking that's going on behind that. I think there's a great opportunity here for AI, right? Because you have a global study, but then you need to do something locally, or even if you just have local studies. There's lots of things that need to be considered. There could be certain relationships between these documents. Obviously, translation needs. There's lots of things there, but the way that we look at our platform and the AI's that we have, I think we can deliver some great value for those types of wave submissions as well in the future. Welcome to AI for Farmer Growth, the podcast from Pioneering Farmer Artificial Intelligence entrepreneur Dr. Andre Bates. This show aims to demystify AI for all those in biofarmer. From start up biotech right through to big farmer, each episode focuses on all things AI and future tech related to help farmer execs to navigate through the and benefit from that sometimes confusing but magical world of AI-powered tools to grow and get real-world results. Today's episode is how AI is transforming regulatory submission and medical writing in biofarmer. AI and automation are transforming regulatory and medical writing in biofarmer by really significantly reducing document creation time while also improving accuracy and compliance. And so tools like these stop co-pilot automate generation of clinical study reports and patient narratives and really cut down the processes that once took weeks or months into minutes. And these AI solutions ensure regulatory compliance, they maintain data security, and they provide audit trails. So they're ultimately accelerating treatment approvals and market access while allowing medical writing teams to really focus on the higher value activities. My guest today is Tim Martin, who is vice president of product, believes ESOP's global product team. And ESOP is a company that specialises in accelerating clinical and regulatory documentation with automation that reduces drafting time and ensures compliance and eliminates a lot of the manual work. So thank you so much for being on the show Tim. Welcome. Great you. Let's start with you telling us a little bit about your background and how you came to be in this role. Yeah, so I've said the majority of my life in Silicon Valley and a technologist and engineer. I left to build things and I've done this for a variety of companies including lots of startup companies. And so I was introduced to EZOP through the chairman and I met the CEO and looked at this as a great opportunity to start leveraging AI to do some pretty cool things for a big farmer today. Yeah, fantastic. So I start with regulatory submission. How is AI specifically changing the timeline for regulatory submission and what kind of bottlenecks does it eliminate from the traditional process? Yes, so the works that we're doing now is really focused on generating the documents in the dossier. So if we take a look at different workflows, there's preclinical, there's clinical, there's CMC, there's pharmacovigilance. So there's a big opportunity here in the dossier and then later. And so we see the opportunity to create massive efficiency just in creating the documents themselves. But I also think that we see the ability for more transformational ROI. So for example, if you automate lots of these documents and take a look at the critical paths through these documents streams, what if you could deliver a draw of weeks or months earlier. What we hear from our customers is that this could mean financially one to five million dollars a day for the farmer company. But more importantly, for patients, they're going to get these therapies sooner. Yeah, exactly. And it's so time consuming, you know, without AI to do all this kind of thing. How do you maintain the human element while increasing the efficiency? Yeah, that's a great question. So we believe that AI should be human centric, right? So humans, AI are better together. And by the way, and this is, I believe, also the guidance that we get from the regulatory agencies, rights, humans are so accountable and responsible for the results of these studies and for these documents that are delivered on behalf of the drugs. So, you know, this is a situation where we can take a lot of the tedious work that medical writers, for example, have to do. They can focus more on the science and the strategy. And you know, there's lots of things in the process that machines can be better at than humans, right? A machines are better at getting data and doing fast analysis there. And so we see lots of kind of opportunities to streamline this make folks more efficient so medical writers can do more with their time. Yeah. Yeah. And how are medical writers adapting to the new AI tools? Are you seeing resistance or enthusiasm for the tools in the field? This is this is also kind of a really important question. So change management in general, right? It's time and effort. If a pharma company puts a lot of effort into change management and looking at AI and that's not just for what we do, but in general, you're going to get better results. And what we've seen as long as there's some strong champions in place to kind of drive that understanding as to what AI can do to help, then you see the adoption happen much more quickly. If you change management programs, a week, then you see more of the pushback and lack of really of understanding of how it can be helpful and really change kind of workloads and kind of their job in general in a much better way, right? Yeah, for sure. And how many hours your clients reporting is saved when they are using these kind of tools for regulatory dossiers, for example? Yeah. So it's hard to put an exact number, but I'll say in general, the expectation for documents, today, the documents that we automate should be north of 70% efficiencies in terms of time savings. And wait a little longer to get better from here. Once again, once we string all the documents, you know, they're them. Yeah, it compounds. Yeah. And what sort of, so you've got pharmaceutical clients, I assume, do you have other types of biopharma clients? We do. So, you know, we've publicly announced that we work with, I believe, it's six of the top 20 pharma companies today. And we also work with some of the smaller biopharma companies as well. Because, you know, we've developed our product to the point where it adds enough value for those folks as well, right? I mean, you look at a big pharma company just automating CSR's is really. Yeah. Yeah. Never mind all the other documents. But when you deal with some of these biopharma companies that don't have as many resources and medical writing and so forth, it's really helpful if we have more complete solution for them and to end. And this is what we're looking to do for them to start efficiencies there too. And beyond just these time savings, what kind of quality improvements come to seeing in their regulatory submissions? Well, this is what I love about AI and automation, right? But only are we affecting, say, the writing, content generation processes, but we can also do some really interesting things that impact processes downstream, right? Because there's there's quality processes that's regulatory review processes, right? After the writing itself, if we do a great job in our tool and help medical writers be much more kind of accurate and complete with respect to delivering a document and the qualities there, the regulatory formatting and structure, all these things that can be tedious are just there. And they just have as a result of the software doing its job, then those downstream processes should take much less time as well. And so this is, you know, something that we're starting to see. And I think this is going to change a much bigger part of the process, not just the writing process itself. Yeah, that's yeah. And what is the balance between the automation and the human oversight and the regulatory writing? That's I'm trying to think about how the best way to describe that is. But I believe if you think about, you know, if you break down the tedious tasks that it takes to create these documents, that, you know, that could be in 30 to 50% of the time, right? I'm in the rough guess, right? So what we're trying to do is kind of build this process in a way where medical writers can do the configuration of the document deal with a few data things early. And so later on, the stress when we say databases are locked in things and they don't have a lot of time and they're being asked to work weekends to deliver things, which will make that much more stress free, right? And if we do a off of this, the medical writers roll changes a little bit in that there's more review and validation, not necessarily they don't have to write everything or deliver all the analysis. AI can do a lot of that that ground. Yeah, they can kind of do the strategic value and thinking to it basically, which I think a key point here is, yeah, writers are great scientists, right? Yes. Yeah. But they don't necessarily want to be technologists. And so part of the job that we need to do is abstract the technology in a way that just makes their lives easier. Absolutely. And that's a common complaint I get with medical writers. I hear, you know, I just don't have time to think I just have to get these things out and I just don't have much time. So yeah, this can save that and give them time for the review and then the value add. Yeah. And I think you have to look at kind of a workflow and to end for them, right? I mean, just having large language model and say, go use this and get more efficient isn't the right answer, right? You know, you build software that looks at a whole process workflow and makes that whole thing easier for you. Yeah, absolutely. And do you integrate with other platforms for that workflow? Yeah, we do as I appreciate that question because kind of it's a natural segue into what we're doing with folks like Viva, for example, something Viva has a great system for content management. And if you look at the process, that can be a starting point for grabbing some of the source documents that we need to do job because that's data into our system. Yeah. And the cool thing is as well, once we do our work, create these documents, we also delivery these documents back into that system, right? So it's a really nice kind of relationship we have with them or one of their AI partners. And, you know, we see that we can really affect a much bigger part of the workflow for these medical writers make their their lives easier in the whole process there. Yeah, that ties into what you're saying before it's the workflow, not just the individual tasks per say. Now if you found the regular tree bodies, have they been receptive to AI generated documentation or have they expressed any concerns that you've heard? Well, I think everybody's a little concerned about AI, but I think the regulatory bodies, there's been kind of limited guidance, but the limited guide has been, you know, this is a tool. Humans are still accountable for the output of these studies or these documents. And so as long as there's accountability there, yeah. It should be fine leveraging AI for writing. Now what's interesting now is I'm starting to see articles that the FDA in the US is starting to take a look at the AI to do the analysis on their side as to whether or not it drugs ready for approval. And this is super interesting because if we do our job and really do a great job of structuring these documents and ring the data in a way that can be used by another AI system on their side, this should also enhance the ability of kind of the FDA to move faster when they use the AI as well. Yeah, they seem to be very open to it. I was talking to someone else recently, not in this space, but they're more in the drug discovery space, but they gave a talk on what they're doing with AI at a conference in Washington. And they said as soon as the talk was over, they were basically mobbed by FDA people wanting to know more and really interested in using it. And so compared with the other agencies around the world, it feels to me that the FDA is actually a little bit more proactive in and more embracing of AI than say the EMA or MHW. Yeah, the EMA is also given positive guidance, right? Just like the FDA. It's good to know. These tools from medical writing, but I think most recently I think you're right, just the things that I'm seeing this week and last week about the FDA's planning using AI is a really strong message, right? That is really good. And looking ahead, which aspects of the regulatory writing do you think will remain firmly in human hands and which will be fully automated? The way I look at that goes back to kind of this idea of interpretation and strategy and scientists doing scientific thinking, right? There's certain things that require deeper thought. And once again, generally, AI and other AI techniques can help with that. But when you're looking at efficacy and endpoints and these things, it requires some key thinking for medical writers. And I think those things will continue to happen, right? We'll make recommendations and they can look at those recommendations and decide how they want to take a look at the data and the trends and the data, but for me, it's really about where they need to work scientifically, not in the air that can just be automated completely, but now they can focus in every document on those sections that requires a little bit more thought. And so do you think these tools, I mean, I feel like they're going to democratize, you know, this whole process for the smaller biopharma companies that don't have as many people and fewer resources potentially. That's our hope, right? Because if there's a sweet of documents that we can support for them and automate for them and make it easy for a small team or even one person to deliver value, then that's absolutely what we want to do for these companies. Yeah. And are there any specific types of errors that human makes that the AI is helping to reduce in regulatory documents? Yeah, so I alluded to this earlier, but, you know, in terms of reviewing data, right? Lots of data, right? Yeah. Things that humans aren't as good as machines, right? So I think kind of this is one area. So this is one area, and I think what's important here, right? If you have errors in the results sections of these documents, it's going to hurt you downstream of the regulatory. Yeah. It's right. And so we really focus on making results sections 100% accurate, right? That's that's really important from our perspective, just to be accuracy, right? But, you know, I think, you know, overall, you know, we'll continue to look at kind of the Pareto diagram of the errors and how these be and say data is number one. And number two, I think it's, it's kind of in these tedious jobs of styling and formatting. Yeah. Do you help with that? Yeah, it's, it's for, we offer, right? So templates that are at ICH driven, transseller H driven, yeah, even create templates for specific customers if they have custom needs. So yeah, this is all handled with our plot. That's really cool. And what about version control during the incorporation of feedback during the iterative development process? Yeah, of course, that has to be there. Clearly, we have an audit trail in our product. Okay, yeah, man. There's lots of things that you need to record to go through this process and, you know, having versions of the documents is an important part of that as well. For sure. And what about for different therapeutic areas that have unique documentation requirements? How are these tools being adapted for those? So clearly we're building a general tool that should work across therapeutic areas, you know, different kinds of studies, different kinds of documents. So when we're looking at the whole dossier mentioned. So and, and clearly the needs for other work strange like CMC and PV are different than the needs for cloud. Yeah, pretty. So we're looking at this thing entire and its entirety. And so this is really important. But I think in general, when you talk about different TAs specifically, there's really things that we can do to kind of build templates and configurations for these different needs and store those. It can be used over. So you don't have to really start from scratch every time. Yeah, that's true. And so what role does AI play in maintaining consistency across different documents within the same submission package? Well, this is the beauty of AI automation, right? Ample, you care about consistency and terminology, right? You do not have consistency in medical writing and how you do that. You care about consistency and of course the results, right? And all these things. So for us, these are all the things that we're tackling. We want consistency, styling, formatting, the way things are written, the way that the data is handled. So all this is really, really important. And this is going to be magical when you take a look at a series of documents or the entire dossier, right? You want to use the same terminology. You want you want everything to look very similar, even from study to study. Yeah, we can do this. Yeah, this is going to be a complete game changer. Yeah. And how might regulatory automation change the skills and expertise that are needed for the next generation of medical writers? Yes. So this is a very interesting question. As I mentioned a bit earlier, I think their job will change the way that they do. The job will change where they'll focus their time more strategically and scientifically. I think that's going to be a big change. I think one of the things that's important for everybody to understand is what AI is good at and what it's not good at. Limitations, because I think that helps folks kind of use the right critical eye when they're engaging. AI, and this is important. If you have business domain knowledge, this is really important, right? Because you have a very good eye in that business domain. But I think this is going to work very, very well together in the future and the change management that we've gone through. Some of our key customers has shown this already. And how are your key customers using it the most? Yeah, I think we're spending lots of time in the clinical space right now. We're starting to get some of these other work streams that I mentioned. And so yeah, we're doing lots of clinical study reports, lots of patient narratives, summaries, working on clinical overviews, all that stuff has to be worked on. We're also working on CMC documents and then the PV area as well. So yeah, starting to make some really nice. Well, lots of different areas. And can you talk a little bit about the security and privacy of the solution? Because clearly if you're working with all those formal companies, they've done quite a lot of checks on your security and your privacy. There wouldn't be working with you. Can you talk a little bit about that? Because I'm sure people are listening. That will be one question that everyone always has because of security privacy and pharma. How do you go about that? How private and confidential is it? Oh, yes. So I mean, this is, of course, the number one of the top priorities when you build this enterprise system like we have, right? You have to absolutely protect all your customer data. And so yeah, there's lots of architectural ways to do that that we employ. We use vendors like AWS. They're very solid and delicious. That we need in terms of the environment. But in general, we go through lots of GXP audits and security audits. And we built this thing to be very, very robust because we can forward any data to ever. No, exactly. That's the thing in pharma. And what about the scientific accuracy? Have you got, well, not even you, just in general, what are the guardrails that companies need to put in place to make sure that the AI generated content is meeting that scientific accuracy standards? Yeah. So there's, there's, there's lots there, right? Yeah. Right down. But you know, there's rules based systems that can be 100% accurate. And we employ those where it makes sense. There's generally eye systems where, you know, that's a little bit more probabilistic there. So you need one guardrails and make sure that you're using that technology in ways that makes sense, right? So, you know, it's just, it's, it's really kind of a fine dance about how you use these technologies to meet the needs of the end users, right? It's still about dealing with pain points. Yes. Now we just have some new technologies to do that. And, you know, to your point, it's a matter of making sure you go back to this fundamental pillar of trust and accuracy. And that's everything. Yeah, that's the first thing that we think. Absolutely. Understand that. And how do the tools handle the challenge of incorporating real-time data updates into documents that are already in progress? Yeah. So it's interesting the way that we're looking at it, as I mentioned earlier, is to really provide a tool that allows you to do lots of configuration and setup early. They kick off of a new document, right? And we want to relieve the late stress when databases get updated. The very end of kind of study activities. And so for us, we should be able to, and everything that we do is take a data update and validate that there's new documents or new data with the medical writer in charge of that workflow. And then update those documents kind of in real time as that happens. And so that it can happen very, very quickly. And then the medical writers can do kind of this final review. Yeah. And in what ways might AI automation actually help address increasing complexity of clinical trials and their documentation? Well, of course, AI can deal with complexity without a problem. It's just a matter of kind of building the system in the right way. But we're already dealing with very complex studies already. I figured you were. We don't see any kind of limits there right now. It's just a matter of understanding all the different inputs that we can get from all these different companies. And we're dealing with the whole set of needs from all the tier one and tier two companies. How does the training and the transition period, when someone comes to you and they start to implement your system? How long does it take to get onboarded and up and running and training everyone so that, you know, if you think of one of your clients that's using it the best, the best use of it. How long did it take them to get to that point? Yeah. So it really depends on their maturity, right? Do they have data? Yes. Where are they in that process? How mature is the medical writing process? How much consistency there is? But we have a really solid customer success team and we can deliver a product almost immediately. Our customer success team can do trials and work with customer data very quickly that they can see the value almost immediately. Then to get something to production, of course, there's a lot of things that need to happen, right? The build, they need to make sure that everything is good on their side for GXP and all these other processes that they care about. But it can usually happen. We can usually get a customer in production in just a couple of months depending on how much time it takes, right? Nice. Very nice. And what impacts has AI automation had on global submissions when there's multiple regulatory frameworks that have to be addressed simultaneously? Yeah. So we keep hearing about wave submissions and the thinking that's going on behind that. Because you have a global study, but then you need to do something locally or even if you just have local studies. So there's lots of things there, but the way that we look at our platform and the AI's that we have, I think we can deliver some great value for those types of wave submissions as well in the future. Yeah, I figured that you would. So if someone listening wanted to get in touch with you, what's the best way for them to reach out to you? Well, if they'd like to learn more about the EZUP products, we have a website, EZUP.com, why SCLP, a little difficult in the spelling, but then I can be found on LinkedIn if somebody wants to reach out. And of course, we're always happy to engage and answer questions about the things that we're doing in the regulatory space. Fantastic. Well, thank you so much for your time today, really, really interesting and clearly very impactful. So, you know, very impressive. Thank you so much for being on the show. Thanks for listening to this episode of AI for Farma Growth. If you have received huge value from the show, or maybe this episode has highlighted how you can use AI in your company, then we would love for you to support the show by leaving us a five-star rating. Please make sure you hit the subscribe or follow button on your podcast app now, to make sure you never miss an episode.