RegTech 2026 in Medical Devices: Capabilities, Constraints & Human Oversight

Yeah. Hey, everyone. Welcome from… sunny… northern, southern Florida. I'm in the car, and so this week, we have the special privilege of having Jose step in as moderator and host. Y'all know Jose. Francesco is gonna be, promoting people, to panelists as we go, so if you'd like to be promoted to panelist, please just accept the promotion. In the meantime, if you want to put in the chat where you're dialing in from, hey Stephanie, hey Daniel, glad to have you here.

And… So, While we're waiting for everyone to join, I wanted to give you some updates about, what's what's coming up. Next week, I don't have the dates in front of me, but I'll paste them into the chat. There's gonna be an ISODIN, ISODIS, 18969 webinar with the chair of the committee, and that's the Clinical Evidence. new standard that's being, developed. I think it'll be super interesting. There are over 900 people already registered, and, so it's gonna be a full house. And yeah, two, two members of the committee that… that…

convene the… the standard. Some really, really interesting thinking going into this. Very, very, lifecycle oriented. And the FDA has already, shown interest in… in the standard, so it's… it's a good sign. Let's see, the following week, we have, Human Factor Strategy, gonna be an awesome webinar, and then we also have part two of the, regulatory. Mdr roadmap, which is, liability. So, I think, that'll be interesting. I don't… I don't want to introduce these guys for 2 more minutes, so I'm just gonna keep filling up airtime. So, next week… oh god, I'm so nervous to say this,

Jose and I are going to be back for an encore performance, introducing our own, new platform that we developed for MLVX, and the MLV 100 for 2026. And so if you get an invitation. from me this week, please join, because it means that you've been selected to be one of the MLV100, the MedTech leading voices to follow on LinkedIn in 2026. And, you won't be sorry if you join. So, because, the new platform is,

AI forward, that's all I'm gonna say. And yeah, it's… it's just amazingly cool. So, we, are gonna get started. As I said, I'm gonna be on… on mute. For today, for the most part. Unless I have prying questions. So, we're looking at… we're here, with Reese and Omer, and I'm gonna, I… Reece, and I met on LinkedIn. This is a beautiful LinkedIn story. We, I got this connection request from this person who used to work at Meta, who was doing reg tech in a startup, and I was like, okay, I'll bite.

And then I reached out to him, and I asked, hey, who are you? And, the conversation took off from there, and we met, and we had a call, and then we had another call, and introduced me to his co-founder, and they are building something really nifty, and you guys are gonna love it. Because we all love change, and I'm excited to… to see the product, to talk to them about it. What I encourage you to think about is, you know, RegTech. What is it? Who's it for? How does it help? You know, and how… how is it gonna, propel our careers forward. That's what we want. And so, on that note, I am going to go on mute. I'm going to turn it over to Jose, who's gonna,

take over from here. Thanks.

Alright, thanks, Sean. So Sean said, so welcome, Reese and Umar. Great to have you guys on. We're excited to hear what you've got to say. Looks like people are still coming in. We're at about 56 people. And as Sean mentioned earlier, if you want to come on as a panelist to ask questions, participate, whatever else, just let us know, and Fraz is going to be adding people as panelists as we go. I think Sean will also be dropping… he made some announcements before 11 a.m. Eastern Time. So if you weren't on yet, he'll be dropping in some information into the chat, so keep an eye on that. If you're camera shy, but you still have questions or comments, drop them in the webinar chat.

Or if you've got questions, feel free to drop them in the Q&A piece of this. And, yeah, without further ado, Risa Neumer, why don't we get a quick intro on yourselves and get rolling.

Cool, sounds great. You want to enter? Oh, sure, I'll go first. Hey everyone, I'm Umar, I'm the engineering brains behind Orchestrate. I've been a software engineer for about 10 years now, started my career at the United Nations, and funny enough, I was working on a humanitarian product there, so that was my first exposure to regulatory environments from the get-go. When you're working with humanitarian aid, you need to log and trace everything That's happening within the systems, and that's where my introduction to regulatory affairs started. Then I moved to Germany, because I fell in love with regulations.

And I've been a software engineer, working there for over 8 years. Most recently had, experience with GDPR and EUI Accessibility Act, and since then, I've just decided to move to London and try to build something for my own, and that's where I'm at, please. Yeah, and so, as Sean sort of alluded to, I think it's quite interesting, the first conversation that Sean and I had, he went and asked the question, you don't see many people go from, like, from Meta to RegTech, and it was the thing where it was sort of like, well, funnily enough. it's more that I, you know, have gone from med tech to meta, and then back to med tech. So, I started off my career as a medical physicist, then did my PhD in,

in, in medical devices, and really got exposure and experience working with companies Pan-African, like with Class 1 and Class 2 devices, as acting as a mentor at the head of innovation at the Global Disability Innovation Hub. And really, you know, one of the things that has always been a persistent barrier to access is, you know, ensuring that products are safe, reliable, up to spec. But also trying to do that without having such a significant, overhead or financial overhead that, you know, can act as a bit of a barrier. So I think what I've always been interested in is obviously, it's great to see innovative products, but I do think that one of the most exciting things is you can think about, okay, how can technology be used to innovate and reduce the barriers to market access? And, you know, that's where I'd sort of say we're called Orchestrate for a reason, because, you know, we know that to get products to market, there are so many moving parts.

And we want to ensure that we're able to, you know, lower the barriers to entry, hold the bar in terms of quality and safety, because obviously it's a critical safety industry. But really, that's our kind of mentality of however we can accelerate innovation and reduce barriers. That's our ultimate mission. It's just that, regs is the area that we're hanging our hat on, I would say.

Awesome, yeah. And Reece, I mean, talking about kind of, like, life cycle of career, am I right, you got your PhD in, like, is it medical device engineering or development, or…

Yeah, so, I think the exact title is something like… PhD in biomedical engineering, computer science, and human-computer interaction. You know, a bit of a mouthful. I think I can remember when I was,

walking on stage and they were reading it out, I think, like, the person who was reading it needed to take a few breaths.

But, I mean, it's one of those areas that I would say it's really been quite interesting, being able to see the medical perspective, from the compci software side of things, tech development, and really trying to, you know, accelerate the product life cycle of things. But then also having the strong, solid background in, you know, physics and science, and knowing, you know, my original My first exposure to medical devices was actually in neonatal brain imaging, so, you know, quite a critical area where You know, everything has to be you know, flawless, exact, but… Yeah, it's been an interesting journey.

Yeah. And just real quick, before we move on, like, how did you end up at Meta from, like, getting a PhD in biomedical engineering and computer science?

Mmm…

Yeah. So…

I think I'll be able to…

Somebody needs to mute.

discipline. Okay.

Alright. Alright, let's see, we'll make sure.

There we go. Alright, cool. I think we're all good. Yeah, so, it was a, it's a funny case, actually. So, being, having expertise in human-computer interaction. One of the things that I was particularly good at was, you know, qualitative research and understanding what the market is trying to do, or needs, what genuine user needs are. And around 2021 was when big tech was in its hiring spree, and I would sort of say that a little bit of luck in terms of they were hiring people with just interesting backgrounds, and then also, you know, luckily I obviously had the… had the right experience, and got into Meta… well, I signed my contract with Facebook, joined Meta, so really interesting point of, change in that company, but worked mostly in advertising, and really working in areas such as

you know, ensuring that privacy policies and the, you know, enforcement of various different emerging regulations from the EU was properly enacted. So kind of a super high burden, like, area where I think there was definitely quite a few months in my career where I'd spend more time working with legal than it would necessarily working with engineering, but, you know, being able to sort of shift gears and understand how you navigate things gracefully when the scale of impact is, you know, a few billion people, and once again, you can't get it wrong.

Yeah, no, it seems like a great background for… for transitioning to RegTech and what you're doing now. So, yeah, why don't you kick us off with the presentation? I know you guys put together some stuff, To… to walk us through that.

Yeah, sure. I mean, so one of the things that we wanted to start with is just the acknowledgement that I think, especially when we look to the changing state of various industries and AI, I think are… almost like we've got it printed out and stuck on our wall, but there's a great meme where it sort of says, you know, I want AI to do my dishes and laundry, I don't want it to do my creative writing and painting. And that's, I think, really the mentality that we're taking. Like, we want to ensure that however we meaningfully integrate artificial intelligence and automation. It's really to kind of…

amplify the expertise, but, you know, minimize the doing the dishes. But I think, really, you know, as I'm sure lots of people on this call know, the state of technology and regtech within medical tech, it's definitely been maturing. I think we're reaching, you know, some interesting points in terms of the companies out there. But I think, really. You know, if we reflect on what these companies do really well right now, it's… things like data management, obviously, tracking registrations, all of that sort of stuff. Auditing, making sure your submissions are… have all the audits and the traces, yes, it's basically structured data management at best. Yeah. Yeah.

And I think, really, what we see as the opportunity is that, as we shift into kind of a more agentic future, not to sound like a too cliche of tech to be using that term, but really it's helping, go beyond just flagging that, hey, here are the problems. It's, like, actually being able to do it in a way that is doing it for you, with you as the key reviewer. And I think that's another thing that I would sort of say in Some other industries, you know. they're going to be taking on board automation and agentic flows where, you know, there is a degree of them devolving the responsibility. That, you know, as I'm sure I don't have to tell anyone on this call, that's not appropriate for the area of med tech. You know, you still obviously need to be able to automate and stand behind every change that's made.

Got it. Yeah, I would love to learn, I mean, so what's… what's the angle that you guys are taking to this?

Yeah, so, and I'll share my screen in a second, but just to cue things up, one of the things that, you know, when we were speaking to reg affairs professionals. It's the common frustration that, you know, I'm spending lots of my life copy and pasting from one document to the other, chasing documents, and actually doing stuff that is, you know, reasonably automatable. But also the tension of the last thing I want is another, you know, a new tool for me to use. So really, we've been building Orchestrate and our software solution on the principle that we want to give you, you know, superpowers in the tools that you already use. So…

whether that's showing up in your word processing, so your Word or Google Docs, whether that's connecting to Teams so that you can kind of have those more on-the-fly requests to manage things, or even in things like, even in things like Outlook, to help you actually, you know, communicate the right things. That's what we're really trying to do, is to provide you that synchronized approach across all of your existing tools. Is there anything you want to add? Yeah, basically, the core is that we meet you where you are most comfortable at. We don't want you learning something new, we don't want you to have 10,000 extra clicks or go to another dashboard or another website. We want to meet you where you're already working at, where you're already comfortable.

And that's… What, we're gonna demonstrate and show right now. Cool, and so just as a quick check, you know, I don't think it's a Zoom call without saying, can you see my screen? I think the only other thing would be if we were on mute, accidentally. But okay, so the… we're going to show two scenarios, two quick scenarios. The first one is how… Orchestrate can help companies that are at that stage where you're looking to expand into different regions. And you've got some documentation and certifications on file. So here, let's say we're expanding into a new market. We're able to see which products, we have in our portfolio. And let's say we were…

looking to expand this specific product. We can see which documents and which registrations we already have. And the key thing here is that because that we already have those technical files and approvals, our system's able to deconstruct them. tag them appropriately, and then figure out how they can meaningfully be reused as much as possible. So say here, if we are looking to expand into Saudi. What we could do is analyze that, you know, here we've got a harmonization pathway, and what we can actually do is reuse the vast majority of things. But as you, you know, you would understand, there's some aspects of localization that would be needed to do for labeling, and you'd need to take some actions to get your, you know, free sales certificate. But what we can do with this information, and that point of going from surfacing

you know, analysis of what needs to change. And when we press Generate Submission, we can actually reuse and reconstruct that information and actually create that document, that's ready to be reviewed. So, here. you know, we've just gone and generated an exemplar document. The intention would be that we know that each company sort of has their own sort of internal house style, so that we can, you know, integrate with that and get that set up as part of onboarding. But… Really, the key thing here is we get you to that 90-95%, then you can kind of go through, see exactly where you need to take further additional actions, and then put the… once you're confident in the document, and you've collaborated and had the back and forth on feedback.

you can actually put it through your formal change management and approval process. So, once again, the principle of we, you know, from just a point of good UX, we want to try and make sure that you're not having to deviate from your existing processes, whether that's reviews, review processes, or, you know, your SOPs. It's really about how do we make sure that we can integrate with that as easily as possible.

Got it, cool. So, so the tool essentially lives in your Google Docs, or your Word, or whatever else, like, you're pulling it up here, and then does it have access to the full design history file, or the full, kind of like a repo of documents or sorts, or how is it… Extracting some of that information that seemed like it popped up.

Yeah, so we're going with a two-pronged approach here. So, one would be indicating with your existing QMS systems, or your existing change management systems, whatever your source of truth is. And we integrate deeply with that to provide this information, or two, we're allowing you, if you're not comfortable… if you're not comfortable with integrating or sharing your QMS information with us. just a drag and drop of the relevant documents that would potentially be needed. Those documents we can already surface for you and tell you exactly which ones are needed to build this information. So all you need to do is just drag and drop them, and

Upload, basically, a window, and then, it… Goes from there and processes all the information to build the the document that you see right now. Yeah. And I think, as well, you know. that principle of we're trying to meet you where you are in terms of your level of comfort and your ability to integrate, you know, across your entire existing tech stack. There is, you know, obviously, as we look to expand over time, you know. we obviously would naturally explore and see where it would be appropriate for us to have, you know, QMS light, I would sort of say, sort of features. Mainly because, you know, when we speak with earlier companies, you know, companies that are maybe registering their first device.

the challenge is usually, okay, so obviously I need to get my 510K registrations or whatever, but, 13485 is kind of the more pressing thing that I need to get in place. So, for the earlier companies, you know, we'll really be providing that offering, but, being… prior… you know, prioritizing and focusing on things. We're focusing more on customers that could benefit from, you know, reusing their existing tech stack rather than, change. Yeah. So, most of the people we've been speaking to have a QMS system, be that, like, an eQMS or, like, a paper-based one. They already have a system in place. Nobody wants to migrate from a well-integrated system and just

completely swap over to a brand new one. So what we're focusing on and trying to do is take away all the grunt work and the tedious, Tedious, boring work that you have to do after, like, in terms of collecting all these documents and preparing all this data, which on average tends to take up, like, a couple of weeks of back and forth between different departments, different people. And all these people are brilliant people with PhDs, and, like, who should be spending their time and their valuable time on, like, making judgment calls, not just going through PDFs. And that's the burden that we want to just eliminate.

Yeah, that's our thought process.

Got it. So, we got about 20 panelists on, and another 65 attendees, And I want to make sure I don't monopolize the questions, because you've got some, like, legitimate regulatory experts, and people in med tech. So, if you've got a question and you're a panelist, you know, feel free to speak up, and then if you've got a question and you're not a panelist, drop it into the Q&A, or if you want to hop on as a panelist, just let us know. But curious… yeah, Michelle, I knew you were gonna have a question, I love it. Yeah.

So this is very interesting, but I was wondering, like, when you look at your competitive landscape, how does this, like, compare to tools like SMVA and, you know, Michael Kania? I can't remember what his product's called, but it puts together technical files.

His product is called MedDevo. E-D-D-E-V-O.

Yeah, so, the thing that I would go and say is really, I think the… the area that we're leaning into is, you know, focusing on that admin side of stuff, and as you say, I know that there are other products on the market that are able to do stuff like that, but really going from some of the magic that we have is going from unstructured to structured, and making it reusable so that you have that kind of full traceability. And then also integrating with our regulatory knowledge graph, which is able to dynamically update and understand how things change. So, you know, I know that other things also, other systems also provide that sort of regulatory, intelligence piece.

The other thing, though, that I would say that we're really trying to grow into is, say, being able to connect across your, kind of, more service-oriented needs. So, for example. We can provide localization services that you can initiate, and, you know, we can be under one master statement of work with your company, and really trying to focus on, you know, building a network of comprehensive partners so that we can help you if you want to expand into a different country. We know that it's not just a simple matter of, well. I just need a certificate that says that we can do that. There's also the logistics, there's the, you know, the more sort of, real-world hard stuff.

So ensuring that we connect, on that side of things is important for us so that we can provide that, you know, value across the value chain.

Here's the questions in the Q&A that just came up.

Yeah, there's some good ones that are popping up, I'm gonna jump to those. Before I jump to those, though, Michelle, I should have said, who are you, and what do you do? Just so other folks…

I'm Michelle Lott, and I own Lean REQA, and we specialize in helping primarily startups and small to mid-sized companies, right-size their regulatory and quality solutions. And we also provide fractional support. Sorry.

Thank you.

And also, I just thought all the interesting questions, the reason they're interested to me is because Tiana, my project manager, is the one that's asking them all.

Oh, perfect.

I are thinking, like, along the same lines.

Awesome. Yeah, and so, I mean, we got a couple… Luke, I know, I mean, you typed it in, but you're on live. Feel free to ask it directly and let us know who you are.

Hi, Luke. I'm an electronic engineer at DEMCON, and FINAPress itself is small, but DEMCON is about 1,000 engineers. Our regular department is about 20 people, so… It's big enough. My question is, so, you, you just said, I just drop a document into it, and it analyzes it. But I can see how it works with textual information, but how do you manage information that's hidden in pictures? Like, there's an annotation, something… this is only valid for Korea, or something you cannot get from the text Around it, but it's immediately, obviously, if you see it in a picture.

Yeah, so on that side, I'll take the question. Yeah, go for it. If you don't mind, please. So, document preparsing and document preprocessing has taken leaps and bounds in the last couple of months. Now we have finally, compliant and industrial-grade document parsers available, which are able to actually go through your images, graphs, and any tables within there, and be able to create them in a parsable enough… in a parsable language that our internal algorithm understands. Funny enough, that's what I've been working on The last couple of days, and going through a bunch of regulatory documents from the FDA, MHRC, MHRA, NMG…

Yeah, he wants, like…

confidence level on everything you expect? Like, say, we're pretty confident on this part, and please look after this part.

Yeah, so… so I think that actually brings in a really important part, you know, like, for example, the… The sort of standards that, you know, you need to have in a product like this is really… human in the loop, expert review. The ability, like, in the back end, to actually surface when… when we're confident and have confidence levels, and say, for example, if it's a very low confidence level. The system needs to be able to self-fix, which, you know, we are able to build that. But at any point when, you know, it requires or is uncertain, it would be the point of, you know, gracefully actually surfacing that to the person so that they can, you know, make that judgment and assessment call.

And the thing that I would say, though, is, like, the point of… the point of… or where we're up to at the moment in terms of technology, and LLMs, I think we are really starting to see a step change in terms of multimodal interpretability. So, for example, you know, with Gemini 3 from Google came out, like, people, people, Gemini's Google system really has managed to, be a step change in being able to interpret visual, and not just textual information. there's a leader for this right now in Llama Index, who are… who've basically released, like, a new version of their document parser, which I've been testing out and playing with it, and it's confidently going through all,

Majority of the images and tables, including graphs, labeling instructions, and is able to convert that into computer-readable text.

Yeah. And maybe for a little bit of context for the folks on the line, you know, where are you guys in the process? I know the product isn't on the market yet. I thought that was one of the interesting things about this, that it's not like, hey, you know, we're here to pitch a product, but really more like, hey, we're working on this very cool product, we know this is an audience that has expertise in this space, and we'd love to get your input on it, because, you know, like you just… kind of brought this up, Luke. Hey, it's going to be really important for you to be able to parse out what's in an image, what's in a drawing, what's in something like that, and incorporate that into it. But yeah, for… if we take a quick pause, like, where are you in this process, and, like, when are you looking to launch?

Yeah, so, I mean, at the moment, we're… we are, working with, you know, we're setting up specific pilot partners. Now. full transparency, we're… we're early stage. And the advantage of that is that, you know. all of, in terms of the direction that we go in terms of feature development, is kind of, to a certain degree, up for grabs. And the reason that we are engaging, you know, audiences and forums such as this, and trying to hook in with the likes of Raps and Topra, is to really ensure that, you know. We're making something that's genuinely useful and can augment are a… as a… as a… as a job role. And…

So, that's one thing that I would say is, like, we've really, you know, coming on calls like this so that we can get that early feedback, and then also, you know, take on specific advisors, so that we can kind of have that, good, good impact, you know, in changing things. But the thing that I would also say is, like, I think both of us have seen, On the tech side of things, in the tech industry side of things, that… the Silicon Valley approach of, like, oh yeah, we're gonna swan in, we're gonna disrupt, we're gonna completely, you know, move fast and break things, yeah, no, that's not this industry, that's not appropriate. And also, I think that, you know.

That, in a way, undermines the value that good, strong regulatory affairs professionals have to the direction of product development and, you know, business development as well. So we want to ensure that, you know, this is the product that, you know, is useful for people on this call, not just for a VC to get excited about.

Right? And so we got a few questions here in the Q&A. You know, one of them is, and probably ties into a little bit of, you know, what you mentioned earlier. There's some tasks in RA and QA that are tedious and, you know, that's not where you want to spend your time, but someone's asking, like, how does it integrate with FDA's E-Star?

Yeah, so, so this is actually, one of the… one of the flows that we're developing is, basically, you'll have your Google Doc or Word document that you'll be able to review, prepare all of your information, but luckily, we are able to also generate a compliant eStar application PDF so that you can then, you know, pass that over, because it would be a bit silly if we were sort of to say, oh, we're reducing admin, and then you get your really nice Google Doc sort of version, and then you have to still copy and paste that over into the eStar template. So, totally hear you that that's, you know, something that we are being aware of and trying to make it so that

We bridge into the formats and submission, formats that are required But depending on the jurisdiction, especially when there's more structured submission processes.

So on the roadmap would be sort of an export to eStar, and then, like, you'd get the E-Star form that's approved by FDA, and it would have already the attachments and the content included?

Yeah. Yes. So I… the reason that I'm pulling this face is because I think the last, few days have been, deep in the schema for how the East Star's constructed, and… We know that it's a… it's a non-trivial task, and… We like a challenge. We like a challenge.

It is a behemoth, isn't it? Yeah. Yeah, yeah, cool. So we got another question here from Tiana, which Michelle was alluding to earlier. It says, does it integrate with any QMS, EQMS, or do you maintain two different libraries of documents?

So, we're planning to… the way we're building this is an API first, for the server side, and with that, we're… we're gonna start with a couple of EQMSs to, integrate with. I think the ones that are on our mind are Greenlight Guru, and there was another… Quadio. Quadio. So we are… these are the two that we're gonna start to integrate with, because these are the two most common and most popular ones. And then as we expand, depending on our customers' needs and, like, how the systems are built, we will integrate with your EQMS system until we say, like, guys, now it's time to use our super awesome, amazing, epic EQMS, which will just audit everything.

and make you, be happy about Rag Tech. I think, though, the point of, like, we… we do recognize that you know, saying, hey, we're gonna help you with admin specifically, and all of that, those sort of tools. Introducing another tool. understandably makes people go, okay, you know, introducing a tool is never frictionless, so that's why we want to ensure that, you know, for example, if you've already got an EQMS, we do as much of the interplay and synchronization with that existing source of truth, because You know, as the, as the, as the question points out, like, the last thing that you want to do is have two golden sources of truth, because that undermines the whole principle. You can only have one.

So the, kind of the long-term vision is you're going to have a means of integrating with an existing source of truth, like a green legur or something, and then an off-ramp in the future, if somebody's ready and trusts your system to say, okay, transition to our platform being the source of truth.

And also, as well, the thing that I would say is, I think, you know, from a cost and complexity side of things, I think, you know, technically, the amount of cost that would have used to go into developing an EQMS, I don't think that those sort of… costs translate, so we know that a common point that we hear, I mean, more so from early-stage companies, is that, you know, EQMSs, even the cheaper ones, are not that cheap. So really, we'd be looking to sort of see how we can, you know, be competitive, for companies that are early stage as well.

Got it. We've got lots of great questions, but maybe we'll take a quick pause to say, if you want, Reese and Umar to contact you guys, because maybe you're interested in piloting this, you're talking to them about the tool. just put in the chat, contact me, or something along those lines. They'll get access to the chat afterward and reach out. if you don't want to put your email directly on there, I mean, I think we've got all your contact information from you guys signing on, but we'll also give you Reese and Umar's contact info, I presume, so that you can reach out. So, we have a question from Rocio. Rocio, I think you're a panelist, so do you want to hop on and ask yourself, or do you want me to read it for you?

I don't think you're on camera, but I do think… I see you on the list of panelists. And maybe while, while Rocio responds, yeah, Tracy, you had a question?

And maybe before you ask your question, just real quickly, who are you and what do you do?

Hi, yeah, so I'm Tracey Lomax, I'm Chief Clinical Quality Officer for Gadax, so,

Yeah, essentially, kind of.

had a lot of experience over the years with, trying to pull submissions. A lot of the time from, not from, you know. Let's say, previous submissions that are trusted sources. So, you know, working more with. what might be sitting in the QMS generally, but certainly always needing to, to kind of Work through a very long, extended period of, like, gap assessment and then remediation. So, kind of interested as to how you kind of see that the tool might be used, to kind of aid that very long, painful process, because, obviously the kind of final point of pulling through from documents are pretty much ready to pull through from.

It's kind of the easiest end of the whole process.

Yeah, and funnily enough, this afternoon, we were, we were just on a call with a prospective partner, who's… doing… registering their first software as a medical device, and they're in the state of… they've got parts of the information. They've got parts of the information in, say, Jira tickets, or that sort of stuff. And really, it's more a case of trying to figure out what is usable, what's not up to spec, and then also surface, like, hey, you're missing this area. You know, to sort of say, we understand, that companies are going to be at varying degrees, it's an entire spectrum from we've got

open a prayer and some stuff scribbled on napkins, to we've got the absolutely most beautiful manicured library of stuff. And we recognize that the real value is, you know, especially as we work with earlier companies. Is our ability to help them know what they're missing and what the delta is in terms of content.

Great. And so, you know, Rocio's question sort of ties into this a little bit, I think. She was asking earlier, like, have you built the logic to detect conflicting information across documents, or is that in the roadmap, at least, where, you know, if you're taking a whole corpus of documents, there's likely to be conflicting information across them, and that's the work that, you know, humans kind of do, and it can be quite tedious. So what's, you know, where do you guys stand on that?

Funny enough, we're working on this, right? It's not part of the roadmap, it is actually the roadmap right now. It's one of the most, core infrastructure elements of our, knowledge graph, or, like, the brain behind the entire thing. And, what we're doing is going through… we're basically creating a causal link and relationship between documents, referencing them through dates and, supersets and subsets to make sure that, hey, if data in document A supersedes data in document… in the form of data in document B, we have that causal link and relationship within our graph network, so that we can provide you an entire response with, hey, this is the latest information, this supersedes the previous information that you were using here, which impacts you XYZ.

And that's one of the general limitations of AI at the moment, that it is good at understanding and inferting things, but it is not good at creating these causal links and references. And that's one of the core elements of our Brain, or knowledge graph, or whatever the fancy new buzzword is. And, that's what we're building right now. Going through this corpus of documents, and creating these causal links, and evaluating them manually, obviously, also, to make sure that none of this is just hallucinations or leads you down to the wrong path. It's tedious, long… monotonous work, but hey, it's a one-time thing that once we've got it done, you guys don't have to worry about that. Yeah. Let us do the boarding job for you.

Excellent. So Eckerd asked such a good question that Luke asked that he ask it, verbally. So, thank you for coming on, Eckerd, and please tell us who you are, what you do, and proceed with your question.

Yeah, I'm kind of semi-retired. I'm doing ISO 13485 audits, or I have been doing quality management system audits, and I worked a lot on automating these, audits by using AI systems, and I always have the problem that people have so many documents, and I think documents are such An outdated habit. What do you guys think about moving, taking the next step, and transforming the input that you get from the companies to structured data? So… That you can get rid of the 90% verbal garbage that is contained in any document, and that you have one single structured.

Source of truth. So, can I… Oh, just, just from a, you know, from a, a human-computer interaction point of view, I think that technically that is obviously the right way, the right direction to be heading in. I think, though, what we have to, you know, when we speak to companies, and especially when there's handoffs between documents, between RA and other parts of the company, that's when, you know, we'd love to sort of say, oh, well, you know, use this specific system, it's a brand new way of, things actually being, you know, a structured, live, kind of semantic map, rather than a sort of. you know, filing cabinet.

That does break down a little bit, though, when you add in the collaboration piece, and until our existing, you know, tools like, like Word, etc, until they change and the way in which we work with each other changes. I think we're going to have to have a degree of accepting that that can be how it is behind the scenes, but hiding as much of that abstraction and complexity of structuring things and having the links between things, but meeting and synchronizing with the current workflows that people are still using.

Right? And Daniel, I know you put a question in the chat, but feel free to say it out loud, and just tell us a little bit about who you are before doing so.

Sure, sure. I'm a former FDA investigator, so you're speaking regulatory affairs activities, which are always linked back to, you know, what investigators do, and then, of course, with the review activities that are performed at the headquarters level. So, your approach, the question that I have is, how do you plan to, sort of. share this wealth of information and this approach that you're proposing to take with the regulators, not only with the FDA, but also with other regulatory authorities, just to make them comfortable, receptive, welcoming of this approach. Because, you know, there's always,

A certain sense of skepticism from the regulators sometimes with anything that comes out new. That was my question.

Great, thank you, Donald.

Yeah, I mean, the FDA themselves recently released their tool, like the ELSA, I think it was called, and… Sorry?

I think it's Eliza, maybe, or…

Right, I couldn't remember the name. And it's been adopted by, what, like, 70 or odd percent of the people, but most of them still say it hallucinates, and it cannot… it can do the bare grunt work, but cannot do the, human judgment and stuff like that. So there's, adoption is increasing for AI, even, like, on a regulatory level, but there's still a lot of skepticism. But our hope is to first build, like, work with, companies such as, like, companies and clients such as yourself to build an established presence and trust within the industry before we start approaching regulators and starting to come up with a more

Yeah, I think that the other thing to say, and I'd use the parallel or changing face of academia and journal submissions as a good example, where we're starting to see, like, as part of the submission process, that, you know, it's asking authors to declare if AI has in any way been involved in the submission process. And I think, you know, that's something that, you know, would be a pattern that we'd welcome welcome from regulators, like, having those You know, proactive ways that acknowledges that The world of, work is quite changing, and actually it's more about appropriately documenting and sharing where, you know, where and how you're using AI. And equally, I think I would love to see from

MHRA, the EU, FDA, actual clarity on, you know, here are the use cases that we think are great for, you know, using this sort of stuff, but just as the FDA say, we aren't using this to judge and, you know, make decisions on whether a document should… whether a submission should be accepted or rejected. Likewise, you shouldn't be outsourcing your judgment. You know, that's… that's an inappropriate use of AI. And that's where we would sort of say, with our system, we're really trying to, you know, make sure that we're heading in that direction, and to this point of, you know, we'd love a conversation as well, you know, we're on this call to kind of make those connections with people.

Crazy yourself.

Let me actually… let me turn the question back to you out of curiosity, having worked with FDA and NFDA, because I… to be honest with you, it's kind of a surprising question to me, because I'm wondering, like, do people have these conversations with FDA? Like, do they go to FDA to say, hey, we've developed this tool, we're following this process, like, we want to educate you on it? Is that… is that a conversation that happens? And also, like. you know, something that Reece mentioned about, or I think it was Wumer, about the adoption of ELISA and AI generally within FDA, and, like, maybe some apprehensions about hallucinations, or whatever the various drawbacks.

But of course, that world is changing so rapidly. I mean, Opus 4.6 dropped this week, you know, Codex 5.3, or GPT 5.3, like. And they're all just getting better and better. So, I'm curious, like, what does that conversation look like with FDA? How do people approach them? And what is the kind of end goal of that from a company standpoint?

Yeah, I think… I think what I'm trying to propose here is that, you know, first, when all of these changes are coming, and now we're trying to bring a certain a new level of communicating with the FDA, because in a certain way, what we do with these approaches is to communicate with the FDA, especially data and information. And FDA's always skeptical in terms of, you know, especially the validity of the data, the legitimacy of the data. It's really, from my perspective, I think it's important that we take that proactive approach to introduce our product to them. To let them know what it is, what it does.

I mean, we send the invite, and we let them know, especially to the review chemists, which become, right now, a brand new generation. Of, of, review people. Probably with a more open mentality than the old generation. So what I'm proposing is that before they ask about it, you tell them. You take charge, you take charge of what you're doing, of how you're commanding, of what is it, the benefit that this product brings to to them. into the patient, right? So that's one, and that's at the headquarters level. But also, when you are visited by an FDA inspector who wants to look at your regulatory data to compare and to analyze and make sure that it matches what you are actually doing on-site.

They also come with a different mentality. And it's also very, you know, skeptical, because their job is really to find out if you're doing something wrong. So, from my perspective, that proactivity starts with the regulators at the headquarters level, where, you know, in a certain way, that information is already now communicated down to the inspectors. So for me, it's just really more about being proactive than reactive.

Yeah, it's fine.

The only point that I just sort of, add as well to this, that, like, say, for example, in provisional discussions that we've had with people in the industry. We… we're hearing back that, say, for example, a common, point of frustration from, say, the MHRA is lack of standardization and predictability in the sort of submissions. You know, I think there is a way of us meaningfully interacting with regulatory Or regulators to actually find out, like, what… What annoys them? What frustrates them? Because, you know, if we can… this is a multi-sided market, you know, it's a classic multi-sided market in that it'd be great if we can automate this, but then if the FDA hates it.

you know, game over. So we really, you know, we totally hear you, and we want to be opinionated, but also collaborative.

Yeah, and it's, you know, when you first asked the question, Daniel, I was thinking about submissions, but I sort of… my mind didn't go into audits. And, you know, there's really… those are two different interactions you have with FDA, and when you submit, the reason I was curious about the question is because, you know, like, what FDA sees is documents. You know, as Reece alluded to it earlier, more and more organizations are starting to ask the question, right? If you submit a proposal to the NIH today, you're actually you're signing something that says that a good chunk of this wasn't generated by AI,

And I haven't seen that with FDA yet, but you might imagine that coming down the pike. But on the audit side, they will surely see what you're using, and they may have questions, and questions about how you validated it, and questions about, like. how do we know that the team that was using it, in fact, did review the documents after they were generated with AI? So, I mean, that's a really fascinating field of all of these AI tools that are being created. I can tell you in the software development world. developers are becoming more and more comfortable shipping code they've never looked at, right? That's kind of like the trend. And it makes you wonder, from a regulatory standpoint, if you're gonna get people who

end up having enough trust in these tools where they're now shipping documents that they haven't read, even though Today, like, you're supposed to be signing off on things that you've, that you, you know, that you've read it, that you agree with it, that everything else, but humans being humans and, and, lazy at times. I can definitely see how, in the audits, this is gonna be really critical.

And I guess just the point that I would add here is this is where, in a way, good UX that doesn't just prioritize, you know, for example. prioritizing giving people, like, a accept-all. Like, that's probably not a pattern that's suitable for this industry. You know, we need to be aware of the points where intentional and thoughtful friction can be integrated into the UX, and is non-negotiable, documented. Audited that sort of says, you know, reset 1649, you know, went and looked at this document, made these changes, and that is shareable as, like, a, you know, kind of a mutable record and log of what happened.

And making sure that, you know, obviously, we can easily export that and share that in ways that can be accessed by auditors. But I think I very much at least. I strongly see that until… society changes, and let's be honest, like, let's look at the legal structure, you know, who's on the hook? It will be the, you know, the person who's responsible for this. Like, we need to ensure that our tech is helping that person to be confident in saying, yeah, I can stamp my my, I can stand behind this document. Rather than, well, I just pressed approval on 150 documents, and let's hope that it's fine. Whereas, like, in software, you know.

You can maybe accept bugs, and you have, like, you know, bug review and QA processes, The parallel I would draw here is, like, the same person who was doing these and going through these documents and going word by word, making sure everything is done, is also just saying, yes, I clicked, I read these terms, and I approved them for a different website for copy consent. So, things that are… that require human judgment, and which actually require you to be conscious and actively present while creating, reading these documents, that won't go away, and automation will never replace that, nor with AI. And it shouldn't.

accountability.

Very comfortable, thank you.

Thanks for the question, Daniel. And Stephanie, I know you had a question as well. If you'd love to come off mute, tell us who you are, and ask away.

I'm Stephanie Grassman, I'm rep… I'm managing the company MedTech Experts, and we offer consulting and support in regulatory affairs, clinical, and QMS. I actually have more than one question, if I can ask them, really. I'm just gonna… rifle them off. First of all, is the source of the regulatory information that you presented here, the frequency of the updating of the regulatory information, and also, are you considering flowcharting the processes, for, you know, regulatory applications, for example, for market access? You know, that you could also integrate. What is your timeline, so that you can sort of work and interact away with the whole team?

You know, so we're not… we're having a real living system rather than a static document-based system. Yeah.

Alright, I think that's about as much as a brain can handle.

For Q&A.

bring up questions. The first two questions are somewhat related, and I think they're really important, right? It's around, like, what is the source that you're drawing from as the agent is compiling some of these documents, right? Is it drawing from FDA guidance? Is it drawing from you know, European, like, and I'm actually curious, just to add a small addendum to that question, you know, when you look at standards, which is, one of the corpuses that you'd… that you draw from, or corpore that you draw from, oftentimes those are, like, copyright protected and that sort of thing. So… so how are you guys handling that, you know, whether it's a RAG system or whatever it is that… where you're capturing this information to make sure that you're providing accurate, documents?

What's the source, and how are you managing that, and how are you making sure it's keeping current?

Yoseh, if I could just get a little bit of background to this. What I noticed with the EIFU for the EU, right, when it came out, it took about 3 weeks before one of the large language models discovered it, and I was constantly checking it as a test to find out how it was renewing the regulatory information, just as background to this.

Yep.

So, the first thing that I would go and say is, to be very explicit, for any standards agencies that might come across this, we are not integrating or making use of any copyrighted material. We'd love to develop that relationship to be able to do that in a way that's appropriate and comfortable, but I just want to, you know, sort of have that very important caveat,

You did work with lawyers for a long time, I can see.

Yeah. What we are doing, though, is, of course, there are excellent, you know, there's excellent open source information that is accessible. that we are able to access, and I'll let you talk a bit of how we Access, and, and… you know, stay current with that, because you know more about that than I do. So, in terms of, like, maintaining and creating the corpus of documents, we are creating a regulatory set for EU, FDA, and the UK. Those are the three regions we're starting with, at least for now. And, as Jeez mentioned, we're using, for now, open source documentation and whatever is publicly available.

But our intention is to work, with regulatory bodies to get access to ISO standards, especially the most important ones, for med tech devices, I can't remember, yeah.

35, 1497, all the…

The usual suspects. The usual suspects, we have them in a list of Notion docs. And, once we, obviously, purchase them, or, like, get a proper official access to them, we're gonna include those to our corpus of documents. And in terms of keeping it up to date, that's something that we're doing manually, and, sort of keeping an eye on, like, whatever addendums and updates are coming out. So, and… and just the thing that I would say there is, right now, we're manually doing that, because, you know, sometimes, sometimes there's a point where actually, you know, you automate and spend, you know, you set up a system that does this, and you spend more time than you actually get back.

We will obviously set up a system that can automatically sync, whether that's based on You know, standards, standard crawler approaches that, you know, is used, but we'd be ideally looking at that from a, a regular. if not daily, then definitely, you know, no more infrequently than weekly, synchronous update. It depends on the corpus we're looking at, right? Like, ISO standards don't change every day or every week, but regulations and some addendums do come up for different sets of documents, so… as, you mentioned, like, the EU AI Act, that is literally a living, breathing document. It changes by the minute.

as the EU regulates and deregulates and regulates more.

Yeah. So, you know, we're running up on time here, got about 5 minutes, and I know people will start dropping off soon, so I'm not going to end the conversation yet, but before we continue, just be sure to drop in the webinar chat if you do want Reese and Uber to contact you, if you're interested in sort of seeing a more exhaustive demo and providing your feedback, I know that they would value just as much you saying, hey, I want to get a license to this. I know from conversations with them, they'll also value you saying, hey, I'm happy to look through it and give you some feedback. Because they're in that critical user feedback phase of this. Just from a timeline standpoint, though, when are you guys thinking that you'll be launching commercially?

So… At the moment, It's looking like we should be able to, start start meaningfully, onboard partners, I'd say kind of mid-March. I think I would love to sort of say, like, hey, you know, sign up and we'll get you on board tomorrow, but I want to also be, you know, pragmatic and realistic. Also, it very much depends on… we need to… meet with you, understand, like, the state of your business, because that changes the timeline. So if someone's got that perfectly manicured, you know, like, beautiful piece, for ski slopes, like, we can probably move a lot faster than if, say, someone's got a… series of napkins with scribbled notes on it, you know? Yeah, so just reach out to us, and based on your situation, we could maybe even launch

start of March, or mid-March, and depends on, like, where we, align, basically.

Alright, great. There's lots of… I had a feeling this was going to be a lively session, and there'd be a lot of questions and interaction. This is a topic that I think a lot of people are interested in, and we're not going to be able to get through all of them, but maybe, just to try to get through one or two more before we sign off. ramald, maybe? I might be mispronouncing that, was asking. could you… could you… could you show us how this tool organizes things like maybe cybersecurity requirements, a topic that's near and dear to my heart, for U.S. and FDA submissions? How do you present the differences and similarities? So, I mean, is this… is the vision for this that ultimately you're feeding information to it, and it's generating all of the documents that are going to be created at the 90-95%, and then somebody's going in and reviewing and changing?

Or, like, what's maybe the last image you want to leave us with in terms of how the tool will ultimately function?

I guess I would say that, the visual metaphor that I'd use is, like, when you go and construct a sandwich at Subway or, you know, other sandwich providers are available. what you do is you actually, you're able to assemble the ingredients in a very specific set way. Now, where the challenge is, is when things are that, like, 90% similar, or, you know, slight deviations, But the, the main thing that we're really focusing on is trying to ensure that we're reusing as much of that information and making a sandwich based on what that industry or that regulation specifically requires. And, you know.

without it sounding too cheesy, like, there's a reason why we're called orchestrate, not Generate. And that's because it is really trying to ensure that we're assembling and making the most of the information that you have. Rather than, sort of. getting, something that might be… a critic might say is AI slop, and we aren't in the industry where that's appropriate.

Alright, great. Since we are coming up on time, Sean, I don't know if you're still with us, but I want to give you a chance, if there's some announcements you want to make. For those of you that came in, like, right on time, or maybe a little bit after, Sean had put some announcements at the top end of the chat, so take a look at that. I know, next Tuesday, February 17th, there's an ISO slash DIS 18969 Webinar that is apparently going to be fired, because there's, like, almost a thousand people who have already registered for it. And, yeah, there's a human factors strategy webinar that's coming up.

as Sean mentioned, next week he and I are gonna be talking about some cool stuff we've been working on for you guys, for the community. And, let's see… yeah, join us next week, make sure you don't miss that. I think it's gonna be pretty cool and fun. Yeah, with that, just want to thank you, Reece and Numer. You know, really interesting. I think there's a lot of exciting developments happening in this field, and you guys seem to have, you know, good personality and approach to it. Like, you're thinking through the lens of Not, hey, we're here to replace our AQA professionals, but really give them the tools to make their lives easier and more effective. So, thank you for your time, thank you all for joining us.

And we'll see you next week.

Thank you so much. Thank you.