Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Service Provider Executive Series #12 - Anita Modi (Peer AI)
86 Plays2 months ago
In this episode, host Steve Gens sits down with Anita Modi, Co-Founder and CEO of Peer AI, for an insightful and forward-looking conversation on the future of AI enhanced authoring. They explore Peer’s founding story, early momentum, benefits of Agentic AI, and what the authoring environment could look like by 2030 — plus what’s ahead for Peer in the coming years!
You can find more about Peer AI by visiting their website at: https://getpeer.ai
Recommended
Transcript
Introduction to Podcast Series
00:00:06
Speaker
Welcome to the Gens and Associates Regulatory Executive Podcast Series, where we explore innovation in the regulatory space.
Meet Anita Modi
00:00:14
Speaker
This is Steve Gens, your host, managing partner, and today I am delighted to be speaking with Anita Modi, co-founder and CEO So welcome, Anita, and been looking forward to this. It's always fun podcast where we can have founder-to-founder discussions. I know we were introduced by a mutual friend, Anton Miak, partner at McKinsey, who we both often talk with. So I'm really glad we got that introduction probably about, I don't know, about six or 12 months ago.
Anita's Background in Healthcare Technology
00:00:42
Speaker
So before we jump into it, if you could introduce yourself and peer AI to the audience, that'll be great.
00:00:49
Speaker
Yeah, well, Steve, thank you for having me here. I'm really excited for this conversation. So my name is Anita Modi. I'm the CEO and co-founder here at Pure AI. In terms of my background, I have been a lifelong operator. I've been a technologist that has focused my entire career in health care.
00:01:06
Speaker
across the industry, really. So on the provider side, public health. But the decade or so prior to starting PEER really focused on building technology for clinical research. Most of that time was actually at a company called Science 37. If you're not familiar, they pioneered the concept of decentralized clinical trials or virtual clinical trials. So bringing trials to patients in their home and I bring this up because it was such a front row seat to innovation in life science, specifically technology-led innovation. And I'd argue perhaps one of the hardest models of innovation out there in this space in the last decade or so.
Founding Peer AI
00:01:41
Speaker
But I had the privilege of serving as the chief transformation and quality officer, leading quality and compliance, cybersecurity transformation and strategy, and worked in my role very closely with our sponsor partners, we call them strategic alliances, all the way from change management to deploying technology and scaling. And I'm really proud of the work that we we did there, driving speed of recruitment, diversity and trials and and retention as well.
00:02:07
Speaker
And given where Science 37 sat in operating and servicing clinical trials, I was really exposed to the regulatory workflows all the way from designing trials, writing these protocols for DCT, which was its own um sort of novel experience through the submission side as well, and started to see how manual fragments fragmented, non-data-driven processes tended to be.
AI-Driven Regulatory Documentation
00:02:31
Speaker
And really with that experience, I launched Peer just about two years ago now to go deeper into how technology can accelerate the journey you know molecule takes through drug development and really get to patients that ultimately need that treatment.
00:02:45
Speaker
um So talking a little bit about Peer AI and and where we operate, Peer AI accelerates cycle times and drug developments by weeks, by months, by delivering AI-driven regulatory documentation.
00:02:56
Speaker
Specifically, we provide a agentic SaaS platform for comprehensive document authoring, allowing life sciences companies to simplify and accelerate their regulatory processes.
00:03:07
Speaker
Now, you've covered this space deeply, Steve, but documentation is a $15 billion dollars market. And you can see that it's the lifeblood of drug development. It really gates progression as you think about the journey a drug takes from preclinical through CMC through clinical and ultimately really took to commercial. And this documentation is what ultimately informs ah regulatory decisions.
00:03:29
Speaker
Teams that are compiling these documents, medical writers, are overburdened with this work. It's such high volume, high complexity, very technical turnaround times expected to be done yesterday. There's actually a shortage of medical writers as well.
00:03:44
Speaker
And these teams tend to be underserved both by services solutions as well as technology.
Agentic AI in Regulatory Processes
00:03:50
Speaker
And automating this work historically has been very tedious, costly, and actually ends up creating new burden or shifting burden in the organization. So I think we'll get into some of the specifics as as we go, but our approach has been to offer an integrated ah approach, one that is driven by agentic AI um with a platform ultimately designed by medical writers as well, which is, I think, positioned us well to transform regulatory document creation. So since our founding just under two years ago now, we've generated over 100 regulatory documents across clinical, non-clinical, CMC, spanning over 10,000 pages or so and counting.
00:04:29
Speaker
And returned over 7,500 hours back to our customers as well. So really showing the value with our you know early biotech and pharma partners of driving efficiency, quality, and reducing cycle time for direct development.
Motivation Behind Peer AI's Founding
00:04:40
Speaker
So really excited to share more with you today. Yeah, likewise. And I um certainly i want to learn a lot more because it's, as we know, it's a changing space. And i think, you know, my early background was with Johnson & Johnson, Janssen specific.
00:04:53
Speaker
My last tenure there was i ran part of ClinOps, but the technology was really specialized. So it was really unique where you had the tech, but also kind of the business process. And this is back in paper CRF, so we won't yeah won't go there. Yeah.
00:05:08
Speaker
It's amazing when technology can be applied and you have that breakthrough. And also, I think in business and anything in life, kind of timing is everything. we go back last 25 years, so we got really good at enterprise content management. We have our templates, we have our style guides.
00:05:26
Speaker
But certainly as a co-founder, you and your colleagues, you had to wake up one day. I mean, I did too, you know, when you started a company. is I have an idea. as this early inception of this could be, you know, we're always looking to incrementally improve, but this is very disruptive. I mean, yeah very disruptive.
00:05:46
Speaker
So why did Peer get founded? I mean, you already shared some of that. And I think the bigger question is what you hope to accomplish in the space? Because you always have to start somewhere as a new business, but you have a longer term vision and goal. So if you could share that with our listeners, that'll be great.
00:06:02
Speaker
Yeah. ah You know, let's let's talk a little bit about kind of the the problem that we saw. i think you're spot on. You know, the industry has been stuck in manual processes despite having decades of incremental tooling.
00:06:15
Speaker
I could argue we've gotten better organized maybe around how we author documents, but it's still rather manual and very error prone. In my seat at Science37, I lived through some of these inefficiencies in trial execution, right? I led the quality and compliance team. So my inbox was kind of a front row seat to what can happen when you have people clicking through systems, sending emails, double data entry. It's not intentional.
00:06:39
Speaker
It's just hard. Peer for us was founded from that pain, the goal to get to ultimately better systems to augment experts or rather be a peer to experts. It's really where we came up with our company name, not so much replace them. And our belief was that AI could genuinely accelerate what they had already been doing so well, right?
Transition to Agentic AI
00:07:01
Speaker
Take the example of a clinical study report. No one wants to spend hours going through 5,000 pages of TFLs here. And so, you know, the automation here of that manual work is what allows experts to focus on the strategic work, right? How do documents align with this like clinical development story you're ultimately trying to tell?
00:07:21
Speaker
Has the clinical data appropriately been framed? And how do you bring the organization along? So much for this team is building that consensus, the collaboration and the time that that that takes aside from sort of the manual repetitive work as well.
00:07:36
Speaker
And so you asked a little bit about why now we're kind of jumping in at this time. It's really what you alluded to. So if I go back to like what's different now versus the past, it really is the advent of generative AI. And, you know, that fundamentally brings just a different level of intelligence of how this work is done.
00:07:55
Speaker
We here at Peer think about the level of ai automation on a spectrum, right? On one end, you have manual static templates, things that are repetitive, rules-based, which I would say are high effort and lower adaptability.
00:08:09
Speaker
As you progress on that spectrum, you get into generative AI, which is faster drafting, but I'd argue limited control and reuse. Still, you need a lot of review, more basic automation,
00:08:21
Speaker
And then the other end of the spectrum, and I think this is where you really start to unlock the potential of technology for this space, this is where a peer plays is agentic intelligence. So what does that mean, right?
00:08:31
Speaker
It sounds like ah a buzzword out there, but agentic intelligence refers to the ability of systems to act autonomously. make decisions, take actions to achieve very specific goals. So again, outcomes and goals rather than specific, very prescriptive asks.
00:08:47
Speaker
And this is the key difference, right? It enables us to be less rigid in how we build and think about more about the outcomes versus the path to output. So press up here, that means things like very intelligent, adaptive templates, dynamically structured content, fully orchestrated agent-driven workflows.
00:09:07
Speaker
um So rather than being kind of straight jacketed by rigid processes, you know, again, we can focus on the outcomes that matter, things like your efficiency, your accuracy, readability, reproducibility, like all these things that ultimately drive quality.
AI's Challenges and Potentials in Regulatory Documents
00:09:20
Speaker
And so, you know, I think being in the space now for two years, it is clear to me, at least, that AI offers immense promise to transform documentation.
00:09:32
Speaker
But I will say that the reality is complex, right? And AI can produce content quickly, but without precise human oversight, the nuanced data interpretation, you know, compliance, readability can suffer, and it can actually compromise critical documents and require time-consuming corrections, right? So I say that because I think getting AI-powered documentation right even with a gentic intelligence, requires having a very clear, simple, and efficient loop between humans and AI and ensuring that AI automation can can scale while meeting these quality requirements. And that's honestly core to how we've been building here at GEAR.
00:10:10
Speaker
So a lot to unpack there. And I will say right off the bat, from a regulatory, we like rigid processes. We have a long history of it. So there's the behavioral change, the process change, the conversation started like three years ago. we have to go from documents to data, but really...
00:10:27
Speaker
what does that mean? But it's it's really, instead of thinking about documents or content, it's kind of components. There's so many, there's thousands of components to a dossier. And at the end of the day, it it all goes through regulatory, right?
00:10:39
Speaker
So that's part of the fascinating change. And I never really understood. So thanks for that, how you came to peer AI. So appear to the experts, because with some of our recent whole studies, we talked about, you know, is it more of a relationship with AI as an assistant, a research assistant? So I think it's Very similar, but that's kind of nice, to peer to the experts. I like that a lot.
00:11:01
Speaker
The one thing, before I get into my next question, I actually have two other points I want to follow up. I really like that step up about about the manual static to the generative AI and then agentic intelligence. So let me try to throw in some time on that. And I think this is, yeah I've used the phrase before, we're we're on a bullet train on this.
Impact of Disruptive Technology
00:11:22
Speaker
is from the manual static templates. Like I said, 25 years of content management and the templates and the generative AI for argument's sake, you know maybe that's started two to three years ago, but that's 25 years.
00:11:35
Speaker
And the iteration from generative AI that was focused primarily on like document types, it's almost like a proof of concept. But two years later, it's already gone to the next stage.
00:11:47
Speaker
So we feel like like every six-month clip, there's dramatic changes. And the other thing that I really liked, and I think it belongs in the identity intelligence bucket that you talked about, is you can get focused. Some folks are focused just on the authoring. Others are like, well, can we use this also? That's stretch our imaginations.
00:12:07
Speaker
yeah to do the QFC compliance on this or ingest all cover letters and compare it to the auto generation about what might be missing. And then finally, the translations kind of in and out too, that if we think about more of that end-to-end process, then some of the results, I mean, they're very
Success Stories and Validation
00:12:25
Speaker
significant. Like with Anton's work at McKinsey, we've referenced it a few times and we have our opinion about the difference between The 2020 data in 2024, as far as the last patient visit database lock, you know, CSR to the filing and the best in class in 2024, I think it was 10.2 weeks for five or six firms.
00:12:46
Speaker
So that's dramatic. I mean, dramatic success. So Before we get into like the future or where you see this, because as ah as co-founders, you probably have a vision of what could be five years out. But before we jump in there, um any early successes you've been getting?
00:13:03
Speaker
Yeah, absolutely. Would love to. Yeah, definitely want to talk about the vision. um You know, it's exciting to have early proof points that validate our approach and particularly one with our more agentic thinking here.
00:13:16
Speaker
so like I said, in the last two years, since we started the company, we've built our AI platform. We've been working with companies from emerging biotech to larger pharma and have demonstrated the value prop of really reducing cycle time. So At a high level, we've seen end-to-end efficiency gains between 55% to 95% less time, and that's end-to-end. So including configuration all the way through authoring, through doing that initial QC as well. And again, driven by this idea of having dynamic configuration, taking in those templates that we have, taking in the style guide, taking in sponsor preference, but using an agentic approach to get those documents
00:13:56
Speaker
authored very quickly as well. But what I think is the most encouraging is the quality piece, right? you know The question is always, well, can you get there faster, but is the quality as good? And so we actually set up our own framework on how to think about quality in a space.
00:14:12
Speaker
I learned early on that you ask 10 people what good looks like, you get 10 different answers. So we went through and interviewed medical writers and boiled it up to criteria such as accuracy, readability, consistency, and completeness.
00:14:26
Speaker
And we've actually had our customers grade the output from the peer platform to say, how does it rank up against this? And we're able to show that we're consistently seeing results that either match or exceed the way that we write today.
00:14:38
Speaker
um So to make this more concrete and share a few examples, you know, I'll anchor us on a simple document. Let's talk about safety narratives. Again, reports that summarize um serious safety events in the course of a trial. We worked with an oncology company, a biotech company, and we authored 56 patient narratives using um their listings files, patient profiles, med watch forms.
00:14:59
Speaker
Traditionally, they spend about six hours per narrative to write that first draft. With Peer, this entire set um took about 16 hours. If you're able to do the math, you're comparing 336 hours with about 16, including the QC process, because really that time was just about QC-ing it as well. And so get a 95% reduction in time.
00:15:21
Speaker
And on a quality perspective, again, we had them grade the output. It was comparable on things like accuracy and completeness and even higher on areas like readability with the assessment that because of that quality, you can actually decrease review time downstream. So again, when you're looking at kind of the end-to-end journey of a drug, really reducing that cycle time.
00:15:41
Speaker
But let's move up to like a more complex document, a clinical study report, which, you know, details the methods, results and conclusions from a clinical trial. We worked with the head of a medical writing and emerging biotech company.
00:15:53
Speaker
This was a six part complex study in patients with narcolepsy or hypersomnia with multiple dosing groups. It turned out to be a Frankenstein study, which was their words, not mine.
00:16:05
Speaker
But in writing this CSR internally, it took them about 40 days to author it over five months. With a pure platform, we did it in less than 17 days.
00:16:16
Speaker
And the customer saw the quality improve compared to what they typically do internally or with their CRO as well. Noting the data accuracy was better even in this sort of complex study, ah but overall just read better as well. So it goes back to idea of can you optimize for readability and consistency and how this story ultimately comes together.
00:16:37
Speaker
But I think what's even more exciting is, you know, what happens when you now run this four or five times on a platform? We're at a place where we can start measuring that compounding value. So we're working with a top 20 pharma right now, and we've shown csr over CSR huge improvements in time. You know, we're already showed, for example, 50% less time between the first CSR we authored and the third CSR.
00:16:59
Speaker
But we're also seeing the quality of the documents improve. And that's the result of this agentic approach, right? We're learning from the human expertise, the verification in each cycle, and you have this continuous improvement to refine clarity, scientific rigor, and compliance and
Building a Strong AI Team
00:17:16
Speaker
alignment. So again, with our North Star being how do we accelerate you know time for interventions to reach patients and really reduce the cycle time, these data points are exciting to show that we're moving in that direction and this technology can drive quality and efficiency as well.
00:17:33
Speaker
Yeah, I think maybe maybe the only other kind of early success I'll share is just personal level for me as a founder. And early success has just been the team that we've built here. You know, from the beginning, it was important to have a team, both of AI engineers and medical writers in-house, given the domain that we're working with. And it's part of our culture to share expertise in and win as a team. I think going back to the point that in this space to thoughtfully apply AI and and get it right, you need this loop of human verification and appropriate human control points to go hand in hand with where you apply yeahi automation. And so medical writers up here actually lead our onboarding, they lead our training, they are product managers in a way um to again, go back to this peer to peer approach of how you think about change management and transformation and having that being led by, you know, ah medical writers as well.
00:18:26
Speaker
Yeah, the people that ultimately do the work. And it it'll be interesting to follow up with you maybe in like the months or quarters to to come. Because as you were going through what you've accomplished with the the CSR, I think a lot of people focus on the writing time. But the the QC time is probably more than the writing time. But where my mind was going, it'd be interesting to track that client over time.
00:18:50
Speaker
And when they submit, does it lead to less agency questions? You know, what does that look like? And I think the other thing, too, for our listeners, that as the thing I'm starting to appreciate about this wave of, you know, disruptive technology we're going through, it's, and I've done a lot of business cases and ROIs in my day, but when you can get compounding value. Yes. um And not not often enough, and we're starting to talk with our kind of regulatory colleagues as they think about efficiency and productivity. And some of the stats that you have are,
00:19:20
Speaker
staggering, but you're creating throughput for an organization, you know a basic concept in manufacturing. You want to increase throughput, but what you're doing is increasing throughput. And I might've mentioned this on one other podcast that we had a client last summer and it was the best regulatory digital strategy.
Long-Term Vision for AI in Regulatory Processes
00:19:38
Speaker
They wanted us, like, we think we got it. Let's have some experts come in and and try to poke holes in it. You know, those are always fun projects. But the thing that we really appreciated with that client, instead of like, hey, we're going to do these generative AI for X, Y, and Z, they they talked and focused on accuracy. So you have like quality with accuracy being a subset.
00:20:00
Speaker
ah But that's critical because unlike other technologies, it works, it doesn't, a workflow, a transactional system. But this is all about the belief in the accuracy. And the accuracy can be the same or better, you know, as a human.
00:20:14
Speaker
And it's weird talking about like human and the third person because it's us, you know, ultimately. it's It's just really impressive at that. Let's go a little bit more into, because these are just pieces of the puzzle that you're talking about, right? So what's the frame or the vision like five years out? Because we're trying to do what's regulatory operations in 2030. And we've done that every five years, you know, since the inception of the company. We're always looking five to seven years out, but...
00:20:39
Speaker
It's a struggle because there's so many, every piece of the puzzle is in play now, you know, as far as opportunities. So there's a lot of excitement, but there's also, yeah, things are changing dramatically. so So what does your crystal ball, or maybe it's not a crystal ball for you, maybe it's very clear five years from now, because, you know, certainly as again, from a founder to founder,
00:21:00
Speaker
Within the first, second or third year of a new thing, you've already taken your best shot, put your stake in. For us, as a small little niche consultancy, we differentiate it by the surveys that we're very well known for. We're actually doing number 47 this fall.
00:21:15
Speaker
But you take your best shot. and And certainly for you, it's the agenda intelligence approach, you know which is a differentiator. But what's that broader, bold vision you have? Yeah.
00:21:26
Speaker
I mean, I'll start by saying this transformation is happening even faster than I expected when we started Peer two years ago. You know, where we are in just conversations and implementation around AI, it looks different, let's say, from the the decentralized clinical trial adoption curve. You know, it is fast.
00:21:43
Speaker
And so I think by 2030 and maybe even earlier, 2028, we're going to see a fundamentally different authoring environment. And I do think this will be driven by what's happening.
00:21:55
Speaker
on the regulatory side right now. So the FDA's AI guidance is here and it's clear they are embracing and actively deploying AI tools themselves like ELSA for submission review.
00:22:08
Speaker
And it sets up what we call it here up here, like the algorithm review challenge. But by 2030, every submission is going to have to satisfy both human reviewers at the FDA and AI systems like ELSA simultaneously,
AI and Human Collaboration
00:22:22
Speaker
right? So it's What does that mean? you know Human reviewers, you and I, are are good at contextual interpretations. right So an example could be when a CSR restates endpoints differently or slightly differently, we would be able to reconcile that. right But for an AI system, these start to come up as inconsistencies and we expect will require you know manual review, add time. It's a different lens of of review.
00:22:48
Speaker
So I think this is actually going to push us towards more rigorous documentation standards, also push us towards even more lean, less repetitive documentation. I think we've already been seeing that over the last decade or so.
00:23:00
Speaker
ah But organizations are going to have to rethink how they design their documentation to meet those sort of dual audience requirements. The FDA's guidance has really emphasized a transparent chain from raw data to regulatory decisions. And I think submissions that meet that standard are going to move even more smoothly through augmented review process. So by 2030, I expect the authoring environment is going to be built around this sort of algorithmic compatibility approach.
00:23:29
Speaker
Documents will need the perfect end-to-end data traceability, link back to source data. I'll see more uniform terminology across protocols, CSRs, INDs, NDAs, BLAs.
00:23:40
Speaker
I think we're going to see more of that consistency. And I predict we'll see more systematic AI-assisted approaches for that reason to get to that area of of compatibility.
00:23:53
Speaker
I think the other dimension here is by 2020, 30, like where do we think the technology will also go? And, you know, I expect the intelligence will get stronger and stronger. You alluded to how quickly even in two years we've gone from just generative AI to more agentic.
00:24:09
Speaker
But I believe that for AI to get actual complex work done, we're still going to require human verification. And so that human AI loop is just going to get more and more critical. We'll have more power with ai but we'll need that to work hand in hand with humans. And i actually think we'll see some cool new UX models ah around that and how that goes handin- hand in hand. That's a lot of what we think about but not necessarily a place of true autonomy. I don't know if we're actually ever going to see that in in this domain.
00:24:38
Speaker
um But to really, again, speed cycle times in drug development, it requires this level of flexibility and expertise that AI will need human help to achieve and work hand in hand with experts. So if you believe this sort of view of where things will go by 2030, I think organizations, the way organizations work will also look different to support this, right?
00:25:00
Speaker
We will move more of the expert knowledge work back to the hands of the domain experts. So again, use AI to enhance that expertise, not replacing them, but working with them again as a peer.
00:25:12
Speaker
I think we'll see some you know new ah content management systems that will be dynamically interconnected. So when data changes, updates propagate through with traceability, this is actually really where AI agents shine. we're We're doing elements of this today and where autonomous activity you know is already starting to prove
Readiness for AI Adoption
00:25:29
Speaker
out. And so think organizations that recognize this shift early are going to have a competitive advantage, faster approvals, fewer ah CRLs.
00:25:39
Speaker
um you know If you think about the sort of model of zero-based redesign, I think the winners are going to be the ones who do that work of rethinking their processes, both the operating model and how teams come together along with the technology um and how it's implemented.
00:25:54
Speaker
and that sort of algorithmic capability will have to be built into their foundations. And so the 2030 Ready organizations are going to ones who see this transformation not as a tooling change, but as a new regulatory language and how they achieve that with humans and technology working hand in
00:26:14
Speaker
And two thoughts on that. I mean, the data we have is already outdated. We published it in February. We had advanced technology and regulatory pulse survey of industry. and it was a nice distribution of the large multinationals, mid-tier, smaller, and also 17 software providers. I think you guys participated in that. And we had ah nine different use cases. So some were generative AI to reg intel, mining of health authority correspondent, dot, dot, dot.
00:26:43
Speaker
But we are really interested in, ah we called it the implementation tipping point, and that's 26, 27. It's very, very clear. And only one company out of the 41 didn't think some of this was viable technology. And I think maybe two or three months after the survey, they probably changed their mind on that.
00:27:00
Speaker
But the other thing, as I was listening to you, that isn't talked enough, and it's a growing conversation about this big change, is different skills. you know like Somebody said, and I wish I could quote it accurately, but it was that the sense of what they were saying is, you're not going to lose your job because of AI. you're goingnna You'll lose your job if you don't have all the AI tool sets in And to have a really ah ancient analogy as my early papers were on a typewriter and then you have this, you know, remember the big it's word or word perfect. That was the big thing. But people were like, well, people are just not going to adopt this electronic technology. And that changed everything, you know, and this is that on steroids. So, but it's the skills I think, and more and more of our clients are talking about.
00:27:45
Speaker
When we look to 2028, 2030, the type of skills and how do we accelerate development, there's generational gaps with all these things, all these complexities when you ever have a ah leap on the ah on the tech and the processes that goes with it.
Peer AI's Goals in Evolving Landscape
00:28:00
Speaker
So usually I like to say, hey we always did the more the longer term. But since this is changing so quickly, really tactically, what should we expect from the peer team in the next, I don't even want to say one to two years, maybe the next six to 12 months?
00:28:17
Speaker
Yeah. ah Before I answer that, I actually want to comment on your note about the skills and organizational readiness. You know, I think now having worked and just met ranges of companies that are different places in their AI adoption journey,
00:28:33
Speaker
I think the ones that have set themselves up really for success are ones that have noted that around really AI literacy and AI fluency and combine the sort of top-down support, right?
00:28:47
Speaker
We're working with companies that have corporate goals right now around automating an IND or a BLA, for example. So really... alignment around it with the bottoms up skill development. And our advice is always just get started.
00:28:59
Speaker
yeah you like Put it in the hands of your employee, whether that's enabling co-pilot or getting an enterprise you know open AI account. And we have actually seen substantial differences when we speak with organizations who have invested in that.
00:29:13
Speaker
just demystifying what AI means, enabling an understanding of where it could be helpful and also where it's not helpful. You know, I think it's not um a magic wand and it can feel like a black box until you actively play in it. So I think you're spot on with even that phrasing, but you won't lose your job to AI. I don't see that happening anytime soon, but I do believe you could be more powerful in your role and this will actually free you up to do other aspects um of your role.
00:29:44
Speaker
right. So your question on this, going back to your question. yeah Yeah. So the next, the next six to 12 months, not one to two years, six to 12 months. Yeah. So, you know, let me step back and talk about the broader vision and then I can bring bring it back to what comes to life in the next couple years.
00:30:00
Speaker
Yeah. Our vision is is bigger than documentation, right? it It's to transform peer AI from today's document platform into tomorrow's regulatory intelligent layer, right? Where every document written, every edit made, every regulatory pattern is learned and made to your next submission smarter, faster, more likely to succeed. Exactly what you were saying about getting that sort of feedback loop.
00:30:23
Speaker
as well. And when we think about the path to get there, there's a couple phases. We're in the first phase, which is and generating you know any document. We've been focused on preclinical, CMC, you know clinical written, INDs, protocols, narratives, CSR. It's really really core of those. And I think what's been powerful about starting there is that every document we author gives us insights into data structure, workflow, like starts to create these connections across functional silos as well.
00:30:53
Speaker
And we're moving from that sort of single document now to more interconnected documents documentation workflows. We're also sort of following data and content workflows, and that's taking us naturally into medical affairs and
Transforming Regulatory Workflows
00:31:04
Speaker
commercial. I think that's going to be part of what you see in the next year or so as well, given the pull from our our customers. So going back to that concept of sort of zero-based redesign and rethinking it versus incremental improvements, that is what this this first phase is, the use of our AI agents to kind of orchestrate new workflows and kind of master that that regulatory language.
00:31:25
Speaker
And i think it sets us up well for the second phase, which is now looking more at end-to-end upstream and downstream workflows, the adjacent workflows that add value to really how this work is done to make that whole process more intelligent.
00:31:38
Speaker
um A good example upstream is supporting more of the Biostats work, right? When we work with our customers and authoring CSRs, one of the source documents is your TFL. It can typically take four to six weeks with multiple cross-functional teams to finalize that. And our initial R&D suggests that you can actually generate that much faster and still enable you know appropriate QC and finalization, but take advantage of AI to do that as well.
00:32:06
Speaker
Another example is even the process of generating key messages out of this data today. You know, it actually sometimes takes days or weeks and you have a team in the tent. We've been testing using our platform for it and you can actually get not just quite close, but start with a first draft for for reflection. And so, again, rethinking a little bit step by step of how this is done.
00:32:24
Speaker
And I think the, you know, and then downstream things into sort of reviews, publishing submissions, we start to create that end to end loop. And I think this is where we start to create that single source of truth and start to kind of streamline more of this work as well. And so for us, we think about the second phase evolving from kind of an on demand document generation tool to more of a continuously running regulatory infrastructure that eliminates some of this fragmentation. And so in the next one to two years, you know you'll see peers start to move into these workflows.
00:32:54
Speaker
But ultimately, sets us up for kind of our third phase, ah which is where we start to become more of the autonomous drug development intelligence layer. Like, I think we're we're early in that, but starting to crack that layer of moving from reactive to proactive. I think the A few examples of that is just a continuous learning flywheel. So every document written, every edit, every piece of regulatory feedback like is already making the system smarter. I think I gave you an example of how we're already seeing that with CSRs, but our agents are getting smarter with every reaction.
00:33:26
Speaker
and those quality improvements start to compound based on regulatory patterns, customer usage, and again, shortening the distance here from your draft to a submission. Start to move into more predictive analytics across a portfolio, right? Understanding content patterns, data relationships to anticipate, like you were saying, regulatory requirements.
00:33:46
Speaker
For example, flagging when a subgroup analyses are are needed as well. And starts to move us again into more proactive things like CRL prevention through pattern recognition, flagging those sort of areas, and ultimately to more of a portfolio level intelligence.
00:34:03
Speaker
Our goal is to move customers beyond that document optimization to understand how changes ripple through your regulatory fabric and know exactly when source data changes, where that needs to propagate and inform the sort of like broader decision-making as well.
Conclusion and Contact Information
00:34:19
Speaker
So for us, going back to our name, you know, we want to be that peer, the trusted advisor that makes those critical development decisions, help optimize trials and navigate this path as well.
00:34:31
Speaker
And it starts with these kind of phased approaches as well. Well, it's certainly very exciting. And it just occurred to me as I was kind of listening with these different phases that, At least when I i started in my Janssen days a long, long time ago, there was always this this thirst, this need for knowledge management. So really what you're talking about is realizing some of the the the goals for decades of knowledge management was always based on what's in people's heads. Now it can be consumed and you have your knowledge management. Maybe there's a knowledge management assistant ah type of thing. Maybe that's coined here on our time. So yeah.
00:35:11
Speaker
Always it's a delight. Great conversation. Excited to learn your progress. So keep us up to date. And we'll probably get up to date through mutual clients too. But I'm sure some of our listening community would love to get a hold of you. so So how should they go about that? Yes. Our website is www.getpeer.ai. You can email us at hello at getpeer.ai or reach out to me, Anita Modi.
00:35:36
Speaker
And for our listeners, you know how to get ahold of us via LinkedIn or through our website. So Anita, thank you so much again. ah Very, very insightful. And from one founder to another, um best of luck. It seems you parachuted in and it's never by accident. It's very, very intentional. And and just having the best team around you you know with your other co-founders that having the right butts in the seat, as somebody ah said, is is just critical on that. So A very exciting time, very exciting company. So thank you very much. Thanks again.