Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Discussing Future‑Proofing Information Architecture in the Age of AI with Generis
64 Plays9 days ago
Welcome back to the Gens & Associates Podcast, in today’s episode host Steve Gens sits down with James Kelleher, Chief Client Officer at Generis, and Max Kelleher, CEO of Generis, for a candid conversation on the future of information architecture in life sciences. Together they unpack how unified data models, governed AI frameworks, and automation are reshaping regulatory operations and accelerating innovation across the industry.
Find out more about Generis at: https://www.caralifesciences.generiscorp.com/
Recommended
Transcript
Introduction to the Podcast Series
00:00:07
Speaker
Welcome to the Gens and Associates Regulatory Executive Innovation Podcast Series, where we explore the latest information strategies in the regulatory space.
Meet the Guests: James and Max from Generis
00:00:16
Speaker
This is Steve Gens, Managing Partner, and I'm happy to be speaking with both James and Max Kellner from Generis. And welcome, gentlemen. This is your second time. It's been about, I think, 12 to 14 months. and boy,
00:00:32
Speaker
Things are changing quick out there in life
Roles and Backgrounds at Generis
00:00:34
Speaker
sciences. So, so happy that you could be with us. So for our listeners, if you could just just do a quick introduction of yourself, maybe some quick words about Generis, and then we can get to it. So take it away.
00:00:48
Speaker
Perfect. and Thanks for inviting us. um's Always great to chat to you, Steve. And it's actually my first time with you because you had a chat with Max and then lee write on that one so you owe me an extra chat. There we go. it's always always good to chat. um So, yes, James Kelleher, I set up Generis back in the day. I've been working in life sciences for years.
00:01:12
Speaker
coming after 30 years and I'm currently Chief Client Officer at Generes, which involves helping to bring customers on board, listening to what they want out of new technologies, new industry regulations, things like that. And then the other half my job is keeping them happy once once they're on board. So yeah, that's me.
00:01:32
Speaker
Yeah. And yeah, great to see you, Steve. It's always good to talk to the life sciences whisperer himself. Hi, it's Keller. I'm CEO at Generis. I've been ceo for two years. I think I'd recently become CEO when we last spoke, Steve, actually.
00:01:48
Speaker
yep um But yeah, I've been with Generis forever. I'm James's son, so I've grown up in and around the company and touched pretty much every area of it on the way up.
00:01:58
Speaker
So I don't have a ah thrilling background in other industries and different dalliances outside of Generis. It's my whole world. So um yeah, that's me. Okay.
00:02:10
Speaker
Oh, great. So let's
Evolution of Information Architecture
00:02:12
Speaker
get started. I know when we're talking about doing this podcast, we're in a really deep discussion just about, it's not so much how technology and everybody's talking about AI, but it's really getting not so much the system architecture, but the information architecture. Like if I go back 35 plus years, James, I stopped counting. um you know As far as life science experience, and as you guys know, I started in Johnson & Johnson you know a long time ago, and there it was mid-range computers, an AS400 from IBM. We had the big mainframes at corporate.
00:02:46
Speaker
Then we got into that client server time, then cloud, and now no it's AI. but It's really, really interesting how things are changing. And if we talk about how systems were actually developed only five to 10 years ago from an information architecture, because that's back in a time when cloud just started. Everybody wants their information connected, be it the information from regulatory to clinical to quality supply chain. So if you're doing kind of a CMC change control, people are expecting all this information is just connected.
Future-Proofing System Architecture
00:03:21
Speaker
But underneath the ah cover, especially with AI, it's very, very different. So with Gineris, you guys have been very successful over, I think we started tracking you, James, about 25 years ago.
00:03:34
Speaker
And you've been through many turns of technology, but always, you know, Generics has always been seen as an innovator in a space. So how do you future-proof the system architecture? I think that's a big, big question today, because even some companies that just came out maybe 10 or 12 years ago, their architecture, people would say their data, which is kind of strange. We might say something's dated 20 or 30 years ago, but how do you future-proof the ah the architecture?
00:04:01
Speaker
Yeah, great question, Steve. And it is weird to sort of say somebody's already dated after such a short period of time. And when I guess technology is changing that fast. But I think Fundamentally, what you had to have been doing five years ago is looking at not just what's the next step forward, but what's the next what's the next whole chapter in this space. And you know things like regulators going towards a much more data-driven approach for regulatory submissions. And
00:04:35
Speaker
companies starting to think of master data as one of their assets and put people in charge of managing that have data stewards and this whole thing. i think that's where the writing was already going up on the wall.
Unified Data Approach
00:04:48
Speaker
And I think that that means anybody who was not five years ago ish five to 10 years ago, already looking at Having a kind of single information, structured information lake, whether that's data or content, anybody who wasn't putting everything in a single place and linking it all together, they were fundamentally kind of just starting off down the wrong path because the days of that siloed approach were very much numbered.
00:05:18
Speaker
So I think that's how you need to look at these things. It's a mixture of, I mean, obviously some of it is reading tea leaves and people get that wrong, but some of it is looking at the overall sort of sense of direction of the industry and and of customers and how they they are looking to inform the design of the software.
00:05:38
Speaker
Yeah, and I think that that's such a small nuance, but it's really, really important because like in our chairs, when we look at all the software providers, you have very few that you know really thought through some of the data that we're going to capture in regulatory. We're also capturing in the QMS and different things. So you're kind of thinking that through and mastering it.
00:06:00
Speaker
within the platform and I can count on one hand with confidence, the amount, you know, yourselves and maybe just a couple of others. So it's I always say that that's more horizontal thinking and less vertical thinking, because if we go back 10 or 15 years ago, it was best to breathe and the cloud came out and then the whole platform. And that's no different because like in manufacturing in the 90s, they took a whole platform approach and then e-clinical, et cetera,
Structured Information and AI
00:06:28
Speaker
et cetera.
00:06:28
Speaker
So Max, anything to add beyond what James has said as far as this you know this information architecture, which is really the bedrock of systems? Yeah, I mean, i think I think that you hit a nail on the head there when you said we're talking about structure and architecture of information as well as structure and architecture of systems. i mean, it's it is two different things, but one begets the other, right? So you can have, i mean, fundamentally, what...
00:07:01
Speaker
What AI needs is access. What even just automation, let's put AI aside for a second, but what automation across a business needs is access to the information real time in order to be able to enable a process with that and be sure that that information is correct.
00:07:23
Speaker
So I don't want to paint pick up a paintbrush and kind of smear across a whole a whole tranche of vendors in the industry, because I think there are definitely some folks out there from an architectural perspective who may be quite...
00:07:39
Speaker
insular but are still able to provide a lot of access to their system. I think the real problem comes with is that data structured in a ridiculously complicated way that is proprietary to that system? And are you able to get rid of all those walls and kind of the structure that you build up within the software?
00:08:03
Speaker
and provide that as a kind of blank slate bucket for other systems to look at. So I think those vendors, our approach aside, which obviously, of course, I think our approach is the best, which is provide the bucket and let them go fish. But I think there are definitely some other vendors out there who say, OK, well, we're going to apply a very intense structure within the system. And we do as well.
00:08:27
Speaker
But we are somehow also going to enable that data to come out of the system with as much ease as possible. And I think that what's really changing, changing I'm going to say changing because it's probably still not changed from some of the largest incumbents in the industry,
00:08:45
Speaker
But there was a defensiveness about the data that is being used within an individual system where you might say, i don't really want people taking data out of our our little network. It's kind of like an Apple ecosystem universe, if you think about it like that.
Challenges with Data Access and AI Governance
00:09:03
Speaker
in that it's easier if everything is the same color working within the same systems that you want it to be and then if you put up barriers to that data going out and being used that's beneficial to you because you get to kind of own the spaces that approach won't work going forwards because if you are running processing if you're running data aggregation platforms I'm sure we'll talk a lot about that. i see But if you're running those sorts of things, if you're putting up barriers for your customers to have their data move in and out freely from the systems, you're going to be the one who loses eventually because the value of what you do as a kind of a system of action or a system of even record underneath that sometimes is decreasing versus this kind of broader enterprise view, in my opinion.
00:09:56
Speaker
I'm just going to pick up on one word you said there, Max, so the changing. The other thing that I think might be changing is this perception of what I was talking about, which is the sort of master data owned by a company. And a lot of people have said, well, you know, it's important on the rubbish out, rubbish in, or the rubbish in, rubbish out route to make sure that your data is good for AI to process. And other people are saying, it doesn't really matter because AI understands the chaos and it can make structure out of that chaos.
00:10:26
Speaker
And so I think the the viewpoint of people is changing on that specific point as well. And I was always brought up to have very clean master data. And yeah when I look at what AI can do, it is surprisingly versatile and able to make sense out of the chaos, which I suppose doesn't really, it's not a great excuse to have leader data in a mess because AI will stop that. But but that is an area that I think is in big flux right at the moment.
00:10:54
Speaker
Yeah, and before we shift, I just want to pick up something were saying, Max, because, you know, kind of that data aggregation platform, we've talked about that, the dApps. And for our listener, if you're not familiar with that term, sometimes people express it as a data fabric, data mesh, data lake. There's different techniques, but you're pulling from multiple systems because, you know, back in the day, even 10 years ago, you just thought about the system and its users, you know, kind of on the consumption where the consumption on these DAP layers and it's pulling from a lot of different functions, you've got to have that open information architecture and that terms mean the same thing. So that's mission critical now. And then let's throw some AI agents over DAPs and transactional systems. Then it gets really, really interesting.
00:11:39
Speaker
So speaking of AI, and James, we'll start with you on this one, you know, too, there's so much discussion and a lot of work going on on AI governance, but, you know, AI in a GXP world, that's a big topic. I'm sure you get that question from your clients. So so what do you tell them?
00:11:58
Speaker
Right. Yeah, it is it is. I mean, everybody that we talk to has got a dozen AI initiatives going on in parallel. There's a lot of, let's say, ah race to the summit internally at customers who the first person to implement a successful approach to a sort of managed governed AI is is going to win the race and rule the roost at that bad customer. So we're going to get pulled into into that. And I guess the viewpoint I have on that is it's really the responsibility of the vendor. So it's our responsibility for providing AI to provide the entire governed framework for that.
00:12:38
Speaker
Why are we any different? Why should you buy our AI versus just uploading things to ChatGPT or Claude or anything? And the answer is because we need to be providing the framework of the tools, the audit trails, the security. Don't tell the user anything they shouldn't know by normal course of business.
00:12:59
Speaker
Track everything they ask, track everything you calculated, track everything you told them. try and achieve that explainability that EMA Annex 22 is all about. i mean, the LLMs, they will do what they are trained to do.
00:13:14
Speaker
It's up to us to to make sure that that happens in a very controlled, structured way. whether that's just pure LLM something like translate or summarize a document or whether it's something which is more of a let's say agentic type of set of tasks so so it i think that's the big responsibility and that's where honestly the but the majority of our focus has gone in providing ai inside the car platform yeah Yeah, thanks for that. It's a big deal. And I had a flashback to when part 11 came out with E-Records, what, two or three decades ago, and there was a panic, you know, and then different philosophies about how far to go. But over time, it's settled into pretty much of an industry standard. And I think it's probably still early days, very important discussions, but it'll work its way out.
AI and Cloud: Integration Complexities
00:14:05
Speaker
So Max, let's change gears again. be interesting to get your perspective that
00:14:10
Speaker
Just kind of the churn of technology as far as if we go back maybe 10 or 12 years with the emergence of cloud-based systems, and now we have the current AI transition.
00:14:23
Speaker
what's What's been more complex to do, the cloud or the ai And I know you guys, you were talking about AI and experimenting before most. Where do you see the biggest benefits for industry in this bullet train that we're all on?
00:14:38
Speaker
So what you think was more difficult, the cloud-based, the AI, and and where are those AI benefits? you know kind of Now, and where do you see them going to, too? Yeah. um I think cloud was, for us, a lot harder because it meant re-architecting everything. And you've got to remember, for Generous at that time, we were also moving from becoming from being a DMS, essentially,
00:15:07
Speaker
into putting an entire database into the platform and dealing with cloud at the same time. So it was a very transformational moment in time for us going full stack there.
00:15:21
Speaker
I think that with AI, as you said, we had, I think if we chosen a different route, which very few companies have chosen, but it hasn't necessarily turned out great for them, but that's the root of like, we are going to become an AI company, which was, you know here's your scoop, was very much on the cards when we were talking about it back in 2018, 2019. And we had these concepts of, I think we had three or four different ideas. We had the librarian who would go back in and improve your data constantly and as a kind of background continuous activity. We had a whole set of, can't remember most of them now, but we had all of these little AI gadgets. I think we were going to brand them.
00:16:04
Speaker
And, you know, we started looking at hiring people and developing the models and and all of that. And then we said, well, we can become an AI company. That will be a massive, massive investment. And we will have to really be chasing that constantly to to keep up with this.
00:16:25
Speaker
Or we can recognize that there will become really great solutions and models out there, basically, either industry agnostic or life science specific. And if we build the framework, it's just what you were talking about. If we build build the framework for governance around those, then we're always going to be at the fighting edge of of AI and bringing that into our customers. So luckily we chose the latter where We weren't building models specifically for life scientists, specifically for our software that kind of put us in a box, but we were creating this framework of governance that that would help all of our customers. And we could always use the latest in the industry.
00:17:10
Speaker
So I think that cloud was was momentous as a shift in the industry. I think that AI is something that will gradually be integrated in one of two ways into all software in our industry, and that is either bolt-on or built-in, right?
00:17:29
Speaker
And so the bolt-on is go out there, find a great platform, tool or work with a partner to bring ai to your current solution and that's usually quite targeted quite focused on an individual feature or something like that or built in whereby you're not doing that yourself but you can go out there and you can leverage the best of that is out there at the moment so we kind of like to think of it like that we've taken the approach where we go what is the best of breed?
00:18:00
Speaker
And to answer the second part of your question earlier is, well, how does that change going forward is we're always going to be able to leverage the best of breed because that's kind of our, and it feels really weird to use a term like best of breed, but it's true. And so we, you know, we we we said the most important thing is our customers are terrified right now and simultaneously really excited about implementing AI because you hear all of these stories, just one from last week where a rental aid rent car agency or something like that had all of their data erased because they they instituted a new agent who disregarded these rules but still had access to everything, right? And took the entire company down basically in one fell swoop.
00:18:51
Speaker
And they were down for 48 hours, something like that. I don't know exactly. but That's terrifying and cannot happen in life sciences, right? So our challenge is how do you just control what access the agents have? How do you control what they can do in the system? How do you control the data that goes to them? And as James said, that's the responsibility of the vendor. That's what makes a difference between is this AI that we can actually sell and operate in this industry or is this AI that...
AI's Transformative Role and Risk Management
00:19:21
Speaker
Sounds really cool, sounds transformative, but could be insanely risky. So, yeah, I think it's about the guardrails. Yeah. Yeah. um And I think one one sort of nuance around what Max was saying was that the reason it was bigger deal for us to go to the cloud than to do AI is because going to the cloud in re-architecting the system we didn't just do what you know maybe would have been an easy route that other vendors might take which is just put the existing stuff in the cloud but actually said how do we architect it how do we prepare at the same time if we're doing all this change let's make sure it's set up for the new year of data being prominent let's make sure it's set up for massive massive scalability and performance and all these things
00:20:11
Speaker
And by doing that, we actually unwittingly, I guess, or maybe maybe wittingly, maybe that's the talent party, built it in a way that was also the right solution for working with AI and leveraging AI to do that. So kind of AI flowed out of that. And that was a big change for us. But that's why AI might have been easier for us than if we hadn't re-architected.
00:20:35
Speaker
so Yeah, I mean, it's kind of a brilliant overview, and especially about like the old term of best of breed is is different now because like a best of breed, like you're not going to go out and write from scratch translation software. What is the best translation software? And recall like with translation software, Greg and I were tracking that for 15 years. and It was just never good enough.
00:20:55
Speaker
But now it is, um and it's really interesting. And I think we're going through that same cycle with the generative AI, because there's a lot of niche providers out there working on the CSR, other things. Some will survive, some won't.
00:21:09
Speaker
I know you guys do a bunch of generative AI, and at some point on the platform, it's going to graduate ah a lot of different document types, document collections to sections of the DOS DA. It's just part of the journey, right? Interesting point on that, Steve. Very interesting points. Max touched on the fact that there were people who were kind of focusing on very specific applications of AI.
00:21:32
Speaker
And we took the approach that we needed to again, just in the spirit of Kara, be flexible, be configurable and whatever. Mm-hmm. And so we'll come up against an RFP where it says, can you generate a clinical protocol? And there'll be other providers whose entire sort of raison d'etre is generating clinical protocols, but but they can't therefore generate a label. And so it comes back to that thing that, i mean, we've been through many cycles like this over the decades and this will resonate with you where Do you provide something to 80%, 85%, 90% good it enough, but works across the board?
00:22:11
Speaker
or do you go very targeted 100%, but it's a really narrow focus? And I mean, I'm not saying there's a right or wrong answer. It's just fundamentally philosophy wise, we chose the sort of broad reach approach. um um And that yeah, that works for us.
00:22:27
Speaker
Yeah, and I think industry is going to have an answer for that in about a year, like 12 to 18 months, because there's so many pilots with what I would say niche providers on this. And that if you think of from a business justification, if you could do, say, generate, and again, it's the translations to and the review cycles. i'm And I hate the term, as you know, human in the loop. We're talking about ourselves in the third person, you know, so...
00:22:51
Speaker
So machine in the loop, right? A type of thing that if you think about it from like a cycle time and business benefit and the cost that if you can generate things good enough so people are just reviewing versus, hey, we can get three or four document types, perfect. That's going to save us a couple of weeks versus hundreds of document types. So it's going to be interesting how that kind of plays out. And like you said, this is no different than things that certain its technologies that are highly specialized, that over time do they get ah
AI as an Enabler for Business Automation
00:23:24
Speaker
evaporated? And most do, because it ends up being in a platform long term, with very few exceptions. but That's usually the rule.
00:23:32
Speaker
So ah one other question for actually for both of you and maybe Max you you can go first is all the conversations AI efficiency governance but you know that word we used to use a lot that we're always innovating so in this like new era how are you helping life sciences organizations innovate? Because you know your company has changed quite a bit over the last seven or eight years. So probably the innovation benefit is has also shifted and I would assume has grown. So maybe Max, starting with you, how do you help life sciences innovate and get your thoughts, James?
00:24:05
Speaker
Yeah, I mean, I think we have to just look at AIs as an enabler of of automation. um And so the innovation essentially comes from how do we affect digital transformation across the entire, let's say, 80% of the business where we have all of these complex business processes with all of their content and all of their data that we need to that we need to regulate.
00:24:36
Speaker
How can we do that across all of it? And the most important thing going forward, when we talk to customers about three years ago, I would have said everybody was obsessed with data governance, which was actually great because good for them. That's set up a really excellent foundation for AI. Pretty useful that they were interested in that at that exact moment.
00:25:00
Speaker
But The reason that that was important is to ensure the consistency and the accuracy upstream and downstream of the data. Right. And again, so today it's still about that across the business. It's not so much about how quickly can I improve my processes within quality or how quickly can I improve my affiliate submission process or so and so forth.
00:25:29
Speaker
It's about We've matured our processes up to a certain point, and now we need to take another step, which is connecting the data, the teams, the processes broader across the entire side of the R&D sphere, if you will, of these companies.
00:25:46
Speaker
And so with really what we're doing with automation and and AI is just a part of that, but is to say, again, it comes back as always to the architecture, but more than that, it comes back to how our solutions are all completely interlinked in a single unified platform. In that we can now build, not only we can provide the governance across all of that, and we know it's going to be consistent because you're not pulling from multiple different sources, but we can also provide now this AI governance framework, which can run and enable automations across all of those different processes whereby, you know, for us, and I'll let James talk to to some of the examples of what we're doing there, but
00:26:31
Speaker
whether it is within the individual use cases, I'm not kind of throwing shade at the individual use cases there and how important they are to automate. But what I am saying is that the next frontier for these companies is the real, is the efficiency of automation of generating content or data for usable formats upstream and downstream across the entire business. So that horizontal play for us, I think,
00:27:00
Speaker
is really important that we are where we are right now. But also I think that we're only just starting to see and realize with our customers the art of the possible within that and how we can transform really, really significantly transform not only just how our users engage with the system, how they manage their critical data,
00:27:26
Speaker
but also organizational structures and how these companies actually work. But yeah, I mean, let's hand over to James for some of those examples are are really exciting. Yeah. um and And they're sort of, they come out, as Max said, they come out of of unexpected
AI Use Cases and SOP Improvement
00:27:43
Speaker
places. So we did, for one customer, set up automations around health authority correspondence and answering those. Very, very sought-after use case. You ingest the regulator's letter. Could you extract the questions? Can you figure out...
00:27:58
Speaker
which ones clinical, which ones are manufacturing, can you search for answers? Can you draft the answers? And by the way, then send them in an email to some, I don't know, hate phrase, some human in the loop. Some human in the loop, yeah.
00:28:13
Speaker
So they can look at. And so that was fine and was working really nicely. And then they said, you know what, our SOP around doing this really is terrible. It's incredibly long. It's very dense reading. People don't understand it. So people have ended up trying to do quick reference guides and trying to sort of do cheat sheets about what this SOP actually means.
00:28:36
Speaker
What can we do about that? So we said, okay, what about asking... the car AI to analyze your SOPs and grade it on readability, on um redundancy and all that. And if it fails, like gets less than, I don't know, 60%, rewrite it, create a short, sharp, very clear to understand SOP. And so these kind of, it's a kind of, what do they call that? an Almost inception thing. As you go inside one process, you find another thing that could be improved and another. And yeah, that I think is the fascinating part that there's not just one or two things you can apply they this to, but really start to reap the rewards across all kinds of different processes. As Mac said, all joined up with that sort of larger picture in mind.
00:29:22
Speaker
Yeah, and I think that's some of the fascination with this. I mean, being in technology, my undergraduate was in technology and kind of how it's solving some complexity aspects. Because like the two of the examples you just mentioned that Health authority, correspondence, and mining, and intake, that's been basically the number one use case in our world since 2020. But it's interesting you bring up the SOPs because we usually cite that as an an example. like This is where we always view AI as an assistant. It can be a research assistant, a writing assistant. But think about like the SOPs in a large multinational, the complexity of their SOPs where
00:30:03
Speaker
Before, you probably would have to bring in a boatload of consultants and spend millions of dollars to analyze them all. And like, how can we make these simpler? Or sometimes in SOPs, there's work instructions, and that's not good. You're changing process.
00:30:17
Speaker
So how can we make them more efficient? But now AI can provide a perspective, not the answer, but a perspective that might be saving people six to nine months. And like you said, it's still really early days on this, but boy,
00:30:31
Speaker
it's It's really, really interesting. So, but... um Yeah, I appreciate the at
Summary and Contact Information
00:30:38
Speaker
the time. I'm going to wrap it up. We've covered a lot of ground from like the evolving information architecture, the big conversation with AI compliance, the importance, because a lot of times we'll say, well, we have an open systems architecture, but is an an open information architecture. And you guys kind of accomplished that by really simply thinking horizontally as others think vertically, or even probably it's more appropriate you think in a hybrid.
00:31:03
Speaker
versus just just vertical, because it's changing. And I know one of your big clients went through a a really, really leading exercise where they went from like non-clinical all the way down to the shop floor and watched the like core data elements. 20 of them, I think they had, 19 or 20.
00:31:20
Speaker
And that's that journey about, like you're saying, it's just... It's just open and information is consumed in very, very different ways and different parties and different technologies versus it's just a system of record and the user relationship. It's really, really has changed.
00:31:37
Speaker
So exciting times. Now, if people want to get a hold of you, what's the best way? Email, maybe off the other website, LinkedIn. So um but what's the best way to get hold of those?
00:31:48
Speaker
any Any which way they find easiest, they can obviously find us on LinkedIn, but also info at generoscorp.com. We'll find a bunch of people who will respond and and pick up any topic.
00:31:59
Speaker
Contact form on our website as well. Yep. Always happy to chat people. All right. You probably have an a AI agent there just waiting, waiting for those correspondence that's coming in. No, they're really people.
00:32:12
Speaker
So. so Yeah, and here, folks, um yeah, just go, we're you know really big on LinkedIn, or if you go to the Gens and Associates page, there's also a contact form to get ahold of us. So thanks again. it was, yeah, I think about 14, 16 months since we talked, Max, and boy, how things are evolving so quickly. Like I think you said very nicely, your client, they're terrified and very excited, you know, and I think that's really the sentiment that we have to be really smart about this, but it's an exciting time. So thanks so much. Thank you.