DCL Learning Series
Modernizing Pharma Dossier Development with AI and Structure
Marianne Calilhanna
Hello, and welcome to the DCL Learning Series. Today's webinar is titled "Modernizing Pharma Dossier Development with AI and Structure: Accuracy, Efficiency, and Compliance from Source to Submission." My name is Marianne Calilhanna. I'm the VP of Marketing here at Data Conversion Laboratory. Before we begin, I want to let this webinar is being recorded, and it will be available in the on-demand section of our website at dataconversionlaboratory.com. You'll get an email reminding you about that. We will absolutely save some time at the end of this conversation to answer any questions you have. But if they come to mind during the conversation, please feel free to submit them via the question dialogue box in your GoToWebinar interface.
I'd like to briefly introduce my company, Data Conversion Laboratory, or DCL, as we are also known. We are the industry-leading XML conversion provider. We offer services that involve structure and content, and data that support our customers' content management and distribution endeavors. Increasingly, we help prepare businesses to be AI-ready. DCL's core mission is transforming content and data into the formats our customers need to be competitive in business. We believe that well-structured content is fundamental to fostering innovation and foundational for your own AI initiatives. We also have Content Rules joining us today. Content Rules specializes in content strategy, structured offering, and new content technologies. They have deep experience in many industries, particularly pharma, medical devices, and other regulated industries such as financial services, manufacturing, and utilities.
We have a lot to cover today, so I'm going to stop talking shortly and introduce my colleague, David Turner. David is DCL's Digital Transformation Consultant. He's an industry veteran in the areas around content management and content structure. David is adept at demonstrating the business benefits of digital transformation and helping organizations identify ROI to gauge their investments in systems, structure, and semantics. We also have Regina Lynn Preciado. Regina is Senior Director of Content Strategy at Content Rules. She's an experienced leader in content strategy, enterprise content management, content reuse and automation, and preparing your content for AI. Regina will be at the DIA Global Meeting in DC in June to talk with you about the content ecosystem of the future. Feel free to email her or connect with her on LinkedIn if you'd like to set up some one-on-one time with her at that event. I'm going to be quiet and turn it over to David and Regina. Thank you both.
David Turner
All right. Well, thank you so much, Marianne, for the kind introduction, and thank you to everybody who is participating here today. Today's webinar is going to be mostly a conversation as opposed to a presentation. I do have some supporting slides, but mostly, we're just going to chat. I'm going to take advantage of the fact that Regina is here, and tap into all of her vast experience,
4:02
and hopefully, bring something different to the table here that you can use, so. Anyway, with that as the backdrop, Regina, I'll turn it to you. We hear a lot about AI in pretty much every webinar. Maybe not just in this industry. Maybe we don't hear as much about structure, but let's just start, and let's just give an introduction about AI and set the table, if you will. In your work with pharma companies, talk about the importance of AI, and what it's doing, and how it's being added, and how we should be thinking about it.
Regina Lynn Preciado
Great. Thanks, David. That's funny. That's such a big question because, of course, AI is everywhere.
David Turner
Yeah.
Regina Lynn Preciado
You cannot avoid it even if you wanted to, and it's in everything from our corporate productivity tools being like, "Hey, I have AI now," "Hey, have a conversation with this document," "Here's an email thread summary," that sort of thing. So to think about or to share what's been very recently quite relevant in my work with submission content, basically, regulatory content is just one example from recent headlines. So the FDA announced recently that it's looking to have companies use AI modeling in lieu of animal studies for monoclonal antibody therapies and a few other things. Of course, there's a lot of work that's gone into that. This is not something they just, in January, said, "Oh, hey, let's do this."
This is a lot of things, but where this headline comes all the way down and impacts my work as a content strategist is, "Okay. How nimble is your content?" because there will have to be new guidelines about what we submit, the data we submit, the narrative we submit with that data or separately. Are we going to submit it in our usual way of the module for non-clinical studies? Are we going to write in the usual sections, what we would have done if it was animal testing, but now, it's the AI modeling? Are we in this period, yes, we are, of needing to respond and provide information in two parallel ways, maybe not for the same drug, but the team is going to work on this and on that, and there's going to be a lot of reporting related around the AI and a lot of reporting related around animal testing, and then what does it mean for global? What is the impact on, "Well, we also need to submit to the EMA, and to Canada, and to the UK, and Australia, and Singapore, and Japan, and everything?" So this brings me back to the other thing that I see a lot of, which is, "Well, can't the AI just do it?"
David Turner
Yeah.
Regina Lynn Preciado
As we're going to get into, it's like, "Well, maybe," but not just you open up AI, and it just works, which everyone on this call and all of your colleagues, that. There's some training and preparation that has to go on. David, I had one more thought on that, which was that McKinsey report you and I were looking at that's from earlier this year where McKinsey – I mean, they've put out a lot, but this one was called "GenAI is a Game-Changer for Biopharma Operations." In that paper, they talk about entry-level use cases for GenAI, novel use cases for – I'm sorry. I said GenAI. I meant AI, including Generative AI.
David Turner
Yeah. Yeah.
7:57
Regina Lynn Preciado
Novel use cases and then frontier use cases, which is a – in the future, we think we'll be able to do this, but the focus of the paper was that novel use cases.
David Turner
Yeah.
Regina Lynn Preciado
In their list of entry-level, no-brainer, of course, we're going to use AI for this, is using Generative AI in what they called automated report writing, and that, to me – these are two things that we've been talking about recently on how automated is your content, how ready for automation is your content. Does that automation need to be ai? Is it just what, three years ago, we considered automation, like automatic formatting and all of that? So there is the picture that's on my mind like this week.
David Turner
Absolutely. I think it's important to say AI can penetrate a lot of different areas, and AI is not just limited to what you said, Generative AI. It feels like in some ways, some people have all of a sudden heard about AI, and they feel like AI hasn't been involved in the process, or they feel like, "Oh, all this content automation stuff is new." But the truth is, is that we've been doing content automation, we've been using AI. It's just this new piece of Generative AI, which, by the way, is super powerful and useful, they feel like that's the only thing. So anyway, I think we want to talk about all of those things. Our goal today is, really, twofold, right? So, number one, we want to inform you about how does GenAI differ from just plain old AI. If AI can't fix all of my problems, what role can it play? If I'm trying to create a dossier, what are the parts that I can automate? How do we need to be thinking about this? So we want to inform about all those things. I think we'll also hit on some other things that are going on, why this is important in light of – I think we talked about it in our prep, some of the regulatory changes that are going to – can you give a quick synopsis of that?
Regina Lynn Preciado
The quickest synopsis of that is regulators are always changing.
David Turner
Yes.
Regina Lynn Preciado
Excuse me. This lingering cough. So, yes. Another thing on my mind is I have a couple of current projects going on in CMC or quality regulatory documents where the chemistry, manufacturing, and control, also known as quality, groups are looking towards the adoption of a new ICH standard. So, for those of you not in that part of – in that silo, ICH, the International Committee on Harmonization has guidelines for all the Module 3 content around chemistry, manufacturing, and controls, and quality. For about 20 years, yeah, we've been working to what is now called Release 1, and there's Release 2 coming out in these guidelines.
Companies will have a few years to really move all the way over to comply with the new guidelines. Part of the difference is these new guidelines are really separating the raw data from the narrative or the highlights. So we're moving information around. You will still be submitting all the same information, it's just going to be in different areas. So, again, like with the AI and animal studies thing, how nimble is your content? Can you easily section out the content and put the information in the right place for that and the right place for maybe something that's ongoing that isn't adopting Release 2? So that's an example of something that's going to hit a lot of regulators because it's coming from the industry and regulator collaboration.
12:01
It's not just EMA is saying everything needs to be more transparent, FDA is saying everything needs to be more data-driven, and different things that the different regulatory bodies are working with.
David Turner
Yeah. All right. So, to our attendees, these are the kinds of things we're wanting to give you some information that you may or may not have. We also, I think, really want to inspire you. When I say that, we're looking at this from the lens – a lot of our preparation we're talking about, specifically dossier development. But I think a lot of these things that we talk about around ways that AI could be used, some of the reasons for using it, some of the structured content, some of the ecosystem things like that can be applied in the other parts of the organization as well, in labeling and in CMC or what have you. Anyway, to get really started now that we set the table here. Before we start talking about – oh, the old way we did it was so bad or whatever. We don't need to talk about the shortcomings yet, but we'll probably hit on it in a second. Let's just start with this. Regina, when you work with a typical organization, how has dossier development been done up until now? What are the typical parts of the traditional dossier development content ecosystem?
Regina Lynn Preciado
Yeah. I think it comes from a centuries-long tradition of thinking and working in documents. So it's a very print-based way of looking at content, and how we've traditionally done it is because what we eventually have to submit at the end is a document. We have authors working in documents and again, print-based. Yes, most people in North America, Europe have moved to digital, so a PDF. I was at a DIA conference a few years ago when – I think it was Brazil was mentioning they had just accepted their last paper submission and had moved everything digital as well, although they meant PDF, not necessarily like database exchange. So it's print-based ideas of publishing documents, and libraries, and journals. Not just the output, the end result is a document, but all the way up from the very, very beginning, the processes are all thinking about documents. Then, finally, it's very –
David Turner
So we're using Microsoft Word here and saving this to a SharePoint system and then ultimately, into the regulatory information management system or some other kind of submission system, something like that?
Regina Lynn Preciado
Yes, and companies will have different RIM systems where sometimes that's where they're storing both the draft, and in progress, and the final documents. There's a few of those different systems that I encounter in my various work. But in general, it is document management, document libraries, even if there isn't literally printing, obviously, there is – labels get printed still, so there is still some printing. It is PDF-optimized, really, for printing, and then the silos, which, of course, any large organization is going to have silos. We're always going to have silos, particularly when you're looking at dossier development with all the people, and all the groups, and all the expertise, and all the knowledge, and all the different groups that come together over years. By come together, I mean contribute to the dossier, contribute to the submission documents as they get finalized and go on.
16:02
And we will always have the silos. So what connects us is going to be the content, and maybe people are looking at their own little piece of that content, except the submission managers who are up here going, "I need thousands of pages of documents, and they all need to be in the right order with all the right cross-linking and cross-references." So, across the board –
David Turner
I put in here these three things you mentioned: print-oriented, document-based, and siloed. I mean, at the end of the day, if we're trying to produce a dossier document, what's the problem with it being print-oriented or document-based? Why is that a problem, and why should we be looking to tools like AI to make this better?
Regina Lynn Preciado
Well, I'm glad you asked. So we are also under tremendous pressure to get faster. Every single pharma company wants to get faster. They want to do more drug R&D in less time. They want to have each drug development get faster and faster. We have in the past, I don't know, 10, 15 years, gone from like, "Ooh, the average time it takes to bring a drug to market is 12 years, and now it's eight years," or something. Don't quote me on that because a lot of people don't really want to say how long it really takes from ideation to marketing approval, but the trouble is working in documents to produce a document at the end is inefficient, there's redundant work, there's a lot of manual high-risk process, like copy and paste, and "Where do you get that information?" "Oh, Diane emails it to me, and then I do this," or maybe it's on the SharePoint. It's not in RIM. It's a lot of people doing the best they can with the tools they have to capture knowledge and exchange information in a very manual way. I think we have optimized as much as we can in many places with other types of process, continuous improvement with some tool integrations, looking for the one tool that does it all, which does not exist, and like with silos, we need to really be thinking of integrations more than one massive system.
David Turner
Yeah.
Regina Lynn Preciado
So then, what happens is it's very hard to meet all the different requirements of all the countries and all the regulators, and even your internal users, and your CROs, and your partners, and if you're pharma partnering with a med device company or one pharma company is developing and another one is manufacturing, like we saw with the collaboration around COVID vaccines, it becomes very, very difficult to find information, discover where there's gaps, and create the information that is needed, share that knowledge, and then produce all the different flavors of almost the same information in the different formats, whether it's the PDF, something for print, a clinical trial registry like ClinicalTrials.gov, for example.
David Turner
Yeah.
Regina Lynn Preciado
The EMA also has a really nice online view of protocols and CSRs, and then you can also download a PDF. So you've got two formats there. So it's very hard to meet all of that when you're not creating in blocks of knowledge upstream.
David Turner
Correct, and I'm thinking about labeling too. When you get to the labeling piece of the organization, well, then you've got – the FDA has SPL that it wants, but other parts of the world still use Word or PDF submissions. Health Canada has been putting out their new scholarly product monograph. You've got FHIR in Europe and Jordan with their own special flavor.
20:00
So, yeah, it definitely is that. But anyway, I think one of the pressures that at least I hear clients talk about is that management says, "Well, hey, we've got AI now so that's going to make all this better."
Regina Lynn Preciado
Well, yeah.
David Turner
If you just add AI, does that really make things better?
Regina Lynn Preciado
Well, it helps us do the things we're already doing faster, at least, I think so. Most of the pilots, and the proof of concepts, and things we've all been – well, not everyone, but the people been experimenting with over the past few years have been very, very carefully curated content. We're taking the best of this, and we're going to run it through. I do, as a writer myself, find that when you have GenAI, it can reduce your time to first draft because you can feed it stuff, and it can get you a first draft, particularly when it knows the rules like, "This is what goes into a protocol. That's what goes into a stability data report." It knows the rules. It puts the medical writer in a different place though because now you're coming in as an editor, as a reviewer, not so much as a writer, and it's very hard to see what's missing in anything. I mean, it's just hard. It's easy to see what's there and go, "That's wrong."
AI can also write a lot of shallow overview sentences that feel right. But when you drill down, you realize you have to put in a lot more information. Now, I say that there are AI tools now that are trained on this specific content. I'm most familiar with some tools trained on clinical protocol, and their first draft renditions, those tools, are much better than a general – I don't even mean open on the web. I mean like your general enterprise like, "Hey, everyone has to use AI. You've got Copilot. You've got the Google one. You've got LAVA. You've got..." whatever. So, yes, I think AI, even in sticking with a traditional content ecosystem, can accelerate the content generation. It can also do things like suggest a little bit better metadata to put on your documents to help with findability. However, we –
David Turner
It sounds like, actually, at the end of the day, you really need a whole new ecosystem.
Marianne Calilhanna
A little bit more.
David Turner
Yeah.
Marianne Calilhanna
Little bit more. Yes.
David Turner
Which is where we're headed here, which is this new modern ecosystem, and it's an ecosystem that I called here an intuitive content ecosystem. The reason I went with this is my thinking when I'm talking about intuitive, when you're putting this together, the organizations that are really doing this now and starting to invest in this now, they're trying to build this new automated and AI-powered ecosystem in a way where the tools and the technologies all really coexist and really interconnect in a way so that the user's experience is a lot more natural, that it's a lot more seamless. One of the knocks on structured content throughout the years has been that "Oh, they've got to learn a new way of working. They've got to learn this new way, this new system."
I think one of the exciting things about some of the AI tools, specifically, these Generative AI tools, is that that with other advances in technologies, it's made these tools a lot easier to use to where writers can actually get excited about using them, and they don't necessarily feel failure. Specifically, what you talked about where you can do GenAI upfront, and then the writer becomes an editor, I've always felt, with structured content, that it's easier to start with an editing experience than it is with a creating experience.
24:04
If you jump into a document that's a structured content management system, and the content is already there, and you're fixing it, and you're not having to worry about this, that, and the other, all the this, that, and the other starts to come in. Whereas if you're going in and you're just trying to create a new one, it's like, "Okay. Wait. Where does this go? Where does this go?" That kind of a thing.
So anyway, I think I'm excited. But anyway, to be intuitive, they also need to move away from the print paradigm and get into more of a digital-first mindset. I think there's this need to move from documents to data. They need to be – lots of connection. I think another big thing is, really, just making it to where you can see the data. If you're trying to allow reuse and connection, it doesn't really help if there's no content visibility. Right? So you want to have content visibility to reduce silence. So anyway, I'm excited about the possibilities, and I think now we've got the different parts of the ecosystem lined up here. So I'm going to hit the next button here and let you talk about the new workflows and what we need to be thinking about with that.
Regina Lynn Preciado
Yeah. Yes, and I want to preface this and remind us in the middle too that what we call these tools today, and they tend to be separate tools, we might be going, "Well, that's the structure content management system. That's the authoring system. That's the AI. That's the database." Whatever we call the tools, whether it's one or many, we're about to go through what the ecosystem is of people, technology, process, and content. However that manifests in the tools, the vendors are working to do new things, us as humans are working to do processes. So this isn't necessarily literally the tools, but it all starts with embracing, "Yay, we love change, a new workflow." That's really where, with Content Rules, we come in a lot, and we realize, over the past 30 years or 31 years almost, how easy it is to recreate your process in your new tool. Then, what you get is all the same things that were problematic before, the whole reasons you went to the new tool, and the promises of your new tool was to be faster, more accurate, less risky –
David Turner
If you've got a print-based workflow and you just try to recreate it, it really as you want.
Regina Lynn Preciado
Yes. Yeah. It's very hard to figure out the new workflow when, I mean, you're steeped, you are an expert in that. So that's one of the ways that I help people is step outside the old way and also reassure everyone you can still very easily produce print or PDF from all this, but the new workflow is where to start. The change management is where to start. So the next piece of this ecosystem is knowing where your data is, and some very obvious examples for this crowd here, of course, are RIM, the Regulatory Information Management system, the quality management system, and then I put in et cetera because when I first said to David, "Hey, we need to list all these sources," it was just this long list of alphabet soup, so, et cetera.
Right now, I'm sure, in your mind, you're like, "Wow, and I work with this, and this, and this, and this." So where's your source data? Then, I hate to say this, but what kind of shape is your data in? Is it? Can you always find the data? Can you always view – back to your point, David, about the visibility of content. Can you always view the data you need while you're writing or working on reviewing whatever you're doing with the content? Do you have it right there? Is it easy to get to? Is it 4,000 steps? Do you have to log in? Do you have single sign-in? All this complication around the data, and not everyone should have write access to the data,
27:56
but it may be a lot of people will need to have more view or read-only access to the data in this new ecosystem than they've had before. So, the data. The next part is what, at Content Rules, we've really started to call Knowledge Capture because in many cases, this is writing. Sometimes it's a chart. Sometimes it's a diagram. Sometimes it's a video. There's different things, but the thing the human knows needs to be captured outside of the human's head.
Do you dictate this into some kind of AI tool that transcribes it, and applies rules, and turns it into a written summary of study design? Maybe, however you capture this knowledge, but this is the knowledge that we as humans have that needs to be put into the ecosystem so that the machines can help us. Generative AI is not going to come up with anything new. It is about ingesting everything and processing all of the data and the knowledge we've given it, and then it can come and follow the rules and say, "Here's the summary. Here's your first draft." So how do you capture that knowledge that isn't already recorded in the system? Very important.
David Turner
I think that is important, especially distinguishing that from your data and your content. Really, that knowledge. One of the things that we've seen fail in the past in implementations of new structured content is the champion of the project starts and gets everybody excited, and then we don't capture all of their knowledge, we don't capture all of the expertise that they have, and then they get hired away somewhere, and they go to another company. Now, the original company has no champion. They've lost all of that knowledge. They've lost all of that data, that knowledge that was in and around the system, and the project starts to go to the wrong direction. So I think capturing the knowledge is important from a GenAI perspective, but it's also important just from a long-term project success, being able to capture those best practices and what people have learned, and to be able to pass that along to other people. Anyway, but I know we got to move on to the next one, which is GenAI itself.
Regina Lynn Preciado
Yeah, and that too, and I was thinking, actually. I hadn't even gone there. I've just learned something excellent out of this webinar, David, because I was thinking of the knowledge capture of the biostatisticians and about the R&D process, but you're right. If you don't have the knowledge of what are the rules, what are the guidelines, how do we do the content ecosystem, how do we maintain it, how do we become nimble enough and stay nimble enough to handle things like ICH M4Q(R2) – just rattle that off –
David Turner
Nice.
Regina Lynn Preciado
Or the changing guidelines from the health authorities around the use of AI in drug R&D and even in generating documents, even in validating that people are following the standard operating procedures, all the different ways these new technologies are popping up in different areas of the company, which leads me to, of course, GenAI, part of this content ecosystem, and I'm specifically talking about Generative AI right here. There's AI, like the AI being used to do modeling of toxicology instead of animal studies. May or may not include GenAI, but for the content, GenAI is where a lot of the medical writers are first seeing it, and a lot of the submission and regulatory document folks are really getting their hands on using useful tools. So GenAI is in this ecosystem, for sure, and it is going to access the source data and the captured knowledge to reduce time to first draft, to check things.
32:00
Now, since 2007, I've worked with natural language processing, which is a type of a NLP that – it's one of the foundations of GenAI, and that has always been about content optimization. "Are you using the right terminology? Do you have the right reading level?" Your informed consent form is typically at a different reading level and level of local language and understanding than the clinical trial protocol that the informed consent form goes with. AI can do some of those, extract the data from the protocol, rewrite it as lay-person-friendly, but you had to capture the knowledge somewhere. You had to have the data somewhere for the GenAI to speed this process up. I know some companies are experimenting with it with their language services providers too around where does AI fit in translations and particularly, in translating to languages that are not just your standard French/ Spanish usuals in North America and Europe, but some of the more rare languages too to reach more populations, so.
So there's that, and then of course, I have to talk about the structured content management system because whether a human writes the whole thing, GenAI some of the thing, your knowledge capture is happening in some other thing and needs to come in, somehow you've got to manage this content. I'm still feeling like structured content management systems, which are designed for this and are very mature, is a good place to store and track, have your traceability, do your data integrations, have the GenAII piece come in. A lot of these tools have a great collaborative review interface now. Maybe they didn't five years ago. But today, they've got some nice – SMEs can come in and easily review. I did have a conversation last week with a CEO of an AI company in the clinical space who says, "I think SCM might be obsolete. I think GenAI or the AI system can do all that." I'm like, "Great." There was no argument that these things need to be done.
Whatever system it's going to be, an SCM, an official SCM system or maybe an AI vendor is reverse engineering some of that content management, and tracking, and reuse, and relationship, and formatting, and assembly, and templating maybe, it's just functions that need to happen. So, again, you have some freedom on the tools. I think the tools – vendors are in this exciting time of figuring out where to go next, but you need some kind of content management system to manage that knowledge that you have captured, add the metadata, provide the context, publish to the XML, the FHIR, the HTML, the PDF, the Word outputs that you need for those documents.
David Turner
Yeah. Go grab the data and bring it into your documents.
Regina Lynn Preciado
Yes.
David Turner
All of that. Yeah.
Regina Lynn Preciado
Yes. Whew, I got all excited. So you can see I'm all attached. I'm like, "I want to keep those." But the tools evolve, and then of course, the Submission Management System, which is you've done all this, and the result, of course, is the submission managers need to get the information in the right place in the right format, which today includes PDF. For some places in the world, it includes print. For many places, it also includes the HTML, or XML, or FHIR, some other digital and database-y type submission information. So your submission management system is going to pull all of this together into – what do I want to say? This is the publishing system, and this is where the publishing system needs to expand beyond print as well.
36:03
Which – these systems also have capabilities to organize and publish other output formats, but I think people – what I see, and this is my last point on this slide, what I see is still this idea that the ultimate goal of all of this is – well, the ultimate goal is drug approval, but the ultimate goal of all of this is produce that package of PDFs for submission. I think we need to change it to be that PDFs is one of our convenient formats that we're going to need for a long time. We'll probably always need it for archiving. But thinking of digital-first, of interactive-first, of personalization-first, let the regulator come in and with metadata, go, "Well, let's see. They need to see this, and this, and this." This other reviewer needs to see that, that and the other thing. Both of them can refer to the PDF if they want the whole deal, but they need their AI systems to be able to pull the information out.
David Turner
Yeah.
Regina Lynn Preciado
Okay. How was that for the nutshell?
David Turner
Well, I like that we talked about – that was great. So I liked that we talked about the people, the process, and the technology. We also talked about a lot of things that don't really fit in the technology, but let's talk about the other piece that you mentioned earlier, the one that when people say, "Oh, hey, I'm going into the project, and I've got to think about people, process, and technology," you also have to think about content, right? Content and why it matters. And the dad joke in me says "Why does content matter? Because content rules." Like, you work for Content Rules and – okay. Anyway.
Regina Lynn Preciado
I get it.
David Turner
I'm still not – dads – there's this great thing on TikTok, these deer hunters in a deer blind. It's like dad is in a deer blind or something, and they go, "Psst." They tell stupid dad jokes back and forth, and I just laugh, and laugh, and laugh. But anyway, that was a sidetrack. Let's talk about why content matters here. I'm going to put up four different things here, and as I put them up, I'll let you talk about them. So talk a little bit about compliance and how content relates to that.
Regina Lynn Preciado
Sure. So we have to provide the right information to the right regulator at the right time and the right place, which in other industries, we might refer to this as personalization. But in this case, we're not necessarily looking – giving the regulatory reviewer a nice experience is a side effect of, "We are complying with the regulation," and it's different in A than it is in B. We've got the right information so it's easy to find, and they can do their jobs quickly, which contributes to the overall acceleration of our process.
David Turner
Easy to find, easy to put together and produce. All right. What about this continuity piece?
Regina Lynn Preciado
Continuity. Preserve and share the enterprise knowledge from ideation through post-marketing. So here is where we talked earlier about silos, and then in the content ecosystem, we talked about it as an ecosystem, different things that depend on other things, and it all kind of has to be cohesive. So as people change jobs, change roles, work on different areas, all that information that today is in our heads and goes past, we're asking each other needs to be captured. The ecosystem can't support that and make it more findable because I think a lot of people write things down, but it's not findable at an enterprise level. With this ecosystem and with AI especially, this is going to get much better, faster.
David Turner
Yeah. I think focusing on the content also does some of this with insights. Go ahead.
Regina Lynn Preciado
Yes. So how do these fancy technologies get faster process, find connections that we as humans would take a long time to find or never make?
40:00
It's because they have access to quality data, quality content, quality documentation. I think we as medical writers and content professionals really see the direct impact of the content on the results of this technology. But I also see that in the leadership and the thought leadership space, people are really thinking of the technology as a technology, which it is. But if there was no good content, what would it give us? I think people understand that story from a data perspective better than they understand the relationship of the content and all of our standards around our content too. For example, the informed consent form or a medical device instructions for use, user guide is going to be written differently than the clinical content and the quality content that went into the R&D. So, insights, and then finding-
David Turner
All right. So the last reason content matters. Go ahead.
Regina Lynn Preciado
Risk management. This comes up a lot. One of the recent examples, of course, is with GXP. So good manufacturing practice, good clinical practice, good documentation practice, whatever your good practices are that you need to look at in terms of managing the risk. So if it's not documented, it didn't happen, and content is extremely important here. Again, we can use these tools to help us get faster, but this is one reason we need to manage the content and not just have our tools create Word documents that we review, and click save as PDF, and send in, which is a very simplistic view of what our big process is today.
David Turner
Absolutely,
Regina Lynn Preciado
Yeah.
David Turner
All right. So I'm going to turn back to the audience here, and I'm going to say we understand that we've given you some things to think about in terms of creating a new ecosystem, and it sounds probably like a pretty big project. You're probably already overwhelmed because anytime you think about an initiative like this or just your management is pushing you to get more AI and to get faster, it feels like an avalanche. So I tried to bring along this graphic here to indicate – if you look on the left, everything is everywhere, and it feels like that when you're doing these kinds of projects.
So, before you can get started, I think the first thing that you've got to do is you've got to get focused and try to bring some of this into line. Think about, "What are the big things that I need to get focused on so that I can then start taking some tangible steps?" So I think the first one of these is stop, take a breath, and find out. Management has, no doubt, given you some guidance. If they haven't, that's an area to start. What are the policies that we have as an organization? Are there particular tools that we're going to use? Do we have a walled garden LLM policy? What are all the various pieces of governance that we have to work in? Regina, I'll turn it to you. What would you add to that?
Regina Lynn Preciado
Yeah. I mean, I think at a very granular level with content, everyone has some very basic things. There's author guidance. Author guidance tells you things like, "In this situation, include this paragraph. In this other situation, include that paragraph."
43:57
Well, decision points like that, that's a policy that sometimes it still takes humans, sometimes a machine could automate that. We have writing standards. We have style guides. We have terminology. All kinds of – we also have, in regulatory documents, a pretty good high-level structure for what information goes where. It has section numbering and so on.
With this content ecosystem of the future, we tend to get a lot more granular within that structure, but we do not start from – we're not reinventing. We're getting more and more granular to manage our pieces of content more like we manage data, and a lot of those decisions have already been made in some companies, you just don't see that that's what it is, and it seems to be done on a document-by-document basis because of document templates. So you can start here by looking across templates, looking across guidelines, looking across training new medical writers, and start looking at what can be programmed and what's not really programmable, and then AI changes the answer to that question, so.
David Turner
Then, that leads you into this next point, which is look for the easiest win. Right?
Regina Lynn Preciado
Yes.
David Turner
Where can you get the best return in terms of your AI and automation?
Regina Lynn Preciado
Yeah.
David Turner
If you can't put in a full content ecosystem on the front-end, where can you start? What can you do? I know you have a couple of ideas on that.
Regina Lynn Preciado
I do. One of the areas is look at how much derivative content is throughout a submission. I mean, the entire Module 2 is derivative content. It is the roll-up summaries. Summary of clinical safety. Rolls up safety information from lots of other documents. Summary of clinical efficacy. Same thing, summary of quality. So summary within a document. You have overviews, introductions, summaries, conclusions. Those kinds of content are a very good starting place even with learning how to train, and tune, and use your GenAI, even if you're not going to the full ecosystem. So that's something to look at and think about even if you're writing them yourself as a human to realize, "Well, this is mostly derivative," and how much time it could save you to have a system that can go through and pull together the key things and roll them up, and then you curate it. So that's one place to start.
David Turner
Excellent, and then I'd say the third area of focus that I would recommend is start thinking about how you can optimize your inputs for best results on the outputs. We can actually do an entire webinar probably on this topic, but I'm just going to leave it at that as an area of focus for now so we can jump into some of the actual tangible next steps. I think once you get focused and now you can start moving forward and thinking about what are the ways that you can get started, that's going to form the basis of putting together a strategy. Now, I put strategy on here, really, as one bucket, but I know when you look at any type of content, you've got to have – what's your reuse strategy? What's your content model? There's all sorts of different faces on that. So what would you recommend other than hiring content rules because I think that's a great idea? What are some thoughts you have around that strategy?
Regina Lynn Preciado
Strategy. We often say strategy, content strategy is a business plan for content, and this goes back to the ecosystem of starting with thinking about your new workflow. Have a plan. We do a phased approach.
47:57
I do notice that in projects where change management and strategy is not a key focus, the projects don't go as well. They take longer. They get rockier. If we start with a strategy, then we always have something to point back to. People often think adopting structure is a heavy lift because at the beginning, we do look for the rules, the policy, the governance. We take all the content apart. We figure out how we're going to put it back together, what could be automated now, what could be automated in the future, and the team that works on that, this could be – it could take as long as you need it to take if you have trouble making decisions or if you start thinking, "Well, we have to automate. This is our one chance to get to perfection," and then you get stuck in that. "We got to do it all at first."
So I have found that some companies that attempted and then abandoned their structure content because it's too hard, they've typically over-engineered a lot of resistance among the writers, which I have seen change because now there's like, "Oh my gosh, now there's AI. I better get with the program on whatever other content automation we have." But because it's not rolled out as well, it's not presented as well, people aren't tied into the business objectives as well, it doesn't feel as easy for one person as pulling up a document, writing something compared to this more collaborative ecosystem of – this is knowledge capture across a lot of people over a lot of years that then gets put together into documents.
David Turner
Yeah.
Regina Lynn Preciado
So strategy is where I advise to start. Have a plan with where you're going to go, and then get into the tools and the technology that's going to manifest your structure.
David Turner
Right. The order makes sense. I mean, the order is critical here because a lot of organizations start looking for the tools and technology first, and they either try to skip the strategy or they don't want to bring in a consultant. I used to work at a place where I said "We really need to encourage people to talk to a consultant," and he said, "Oh, if they need a consultant, then our tool is just too hard." That's not true. It's just that the consultants who help you with the strategy can help you to pick the right tools. Remember, we're trying to get these tools to talk and to work together, and so – one particular submission system might work really well with another kind of system over here. You need to think about – cut all those things.
So tools and technology, I think, are next, and then you have to think about, "How do we get this content structured?" If you haven't worked with structured content before which – we haven't even really hit on the whole idea of how structured content really fuels AI and really makes your AI that much more accurate, but that's another webinar we'll do. Breaking your content into little chunks starting with documents can be a challenge. It's not simply, "Oh, hey, we're going to take this Word document and convert it to this." You've got to actually put some smarts into it. You got to break it down consistently. You've got to look at areas of potential reuse. It just so happens that's what DCL does. We convert content from Word into structured formats and tag that.
Then, we also have a tool that helps to identify, "Hey, here's reusable content across all of these different pieces," so that you can really start making smarter use of that. While I would love for companies to come to us at the beginning, it really does not make sense until they have an idea of where they're going and what tools they want to use. But sometimes we get involved along the way just to help be thinking about that, and we do work closely with Regina. Then, I think the last thing here in terms of getting started, start looking for some industry groups that you could be a part of. What's going on with the standards?
52:00
What's going on with the submissions? Who's talking about structured content? I know DIA has some different communities for this. I'd love to hear your thoughts, your questions, or – yeah. Sorry. Your thoughts, your ideas from the community out there. If you want to put things into the Q&A, "Hey, I'm a part of this group," or whatever, if you want to email us afterwards, that kind of thing as well. But I'm going to go ahead in here because we're getting close to the end. I do want to make sure we have time for questions. But before we jump into those questions, I want to ask you this, Regina. Any final thoughts about what we've talked about today and moving forward, et cetera, before we bring Marianne back on?
Regina Lynn Preciado
Yes. So I want to reassure that this can be done. Other industries have adopted structure and are now in a great place for their GenAI roll-outs because their content is already chunked and contextualized with metadata and relationships. Some of them have full-on knowledge graphs already. Some of them hadn't done that yet, but are working on it. So, in pharma, we've been slower, and a lot of this has to do with just approaching it like a new technology and an idea that you have to write completely differently. Everything is going to be completely different, and now you have to write for the machine, which is not necessarily with structure, although there are ways to optimize so that AI can really automate and work with your content as well as the automated formatting, and the translation management, and all that stuff we get with structure.
So I just want to say it can be done if you have a plan, and you see it through, and really understanding that there's a perception about what a heavy lift this is, but the heavy lift doesn't have to be that heavy, and it's the one-time thing. Once you're set up, it becomes a little bit more like a factory. It's not as handcrafted as it used to be, but that's what gets us to our business objectives of getting more efficient, less risk, faster, and ultimately, getting medicines to patients faster, which is our main purpose here. It is not writing the documentation, so.
David Turner
It can be done incrementally too, right?
Regina Lynn Preciado
Yeah, it should be.
David Turner
I mean, it's one of those things where some people try to just do it all at once. "Oh, hey. Well, this is the year we've got budget." Well, I think you've got to approach it more carefully, and I think this again is another real use case to use a good consultant in this area, someone to communicate with management, "Here's where we're trying to get, here's why we're trying to do this much and not that much, and here's how AI is going to be communicated," and then making sure that you just are communicating and working that all the way through the process. All right. Well, that brings us here to the 5-minute mark. So, oh, there's Marianne. She's back. What questions do we have to answer today?
Marianne Calilhanna
Okay. We have a couple questions here. Let's see if we can get through all of them. Okay. Someone asked "While we try to automate content generation, are there any discussions with any of the health authorities regarding the future of content? The ECDT, the Electronic Common Technical Document, is already 20 years old. That's the mechanism to share information with the agencies. Do we really need to write the introduction 76 times?"
Regina Lynn Preciado
Who is that? Can we go get coffee? [Laughs]
David Turner
Yeah.
Regina Lynn Preciado
Wouldn't it be great if we could provide all the information one time?
56:02
A bunch of chunks, single source of truth, and the system on the other side could pull together the thing that particular reviewer needs to see at that particular time? So everything is there, and we are not needing to go, well, this audience might need it written like that, and this audience might need it here. The safety people want all the safety pulled out of everything and put here, because that is something that is consistent, repetitive, manual, and that's the kind of thing that can be automated and in fact, doesn't even need AI to do it, but AI is our emerging speedy way of doing things. So, yes, those conversations are going on. There are some industry groups talking about it. If you can attend some of those conferences, like I tend to go to DIA, the Drug Information Association conferences, but there are lots of others, which, again, to David's point, if you want to email me and let me know where else I should go.
David Turner
Yeah.
Regina Lynn Preciado
Those conversations are happening in those conferences that bring regulators, and sponsors, and CROs, and manufacturers together. I know there are some industry groups talking about it. I am sorry. I have just completely blanked. There's something like "the Life Sciences Innovation Group."
Marianne Calilhanna
That's it. It's Life Sciences Innovation Group.
Regina Lynn Preciado
Thank you. So anyway, go ahead, Marianne.
Marianne Calilhanna
All right. David, you referred to content reuse a couple times. My company has a mountain of documents. Where do I even start around that topic, around content reuse?
David Turner
Yeah. Well, I think there's a couple of things. Some start by just trying to set some sort of a baseline for content older than this. Maybe we don't need to have content older than that. Others will start by just trying to get it organized. "Hey, can we try to get it all in this same place?" One of the other things we recommend is running an analysis of where the potential reuse might be. We have this tool called Harmonizer that we use, and we can basically upload all of your Word documents, and PDF documents, and whatever other format you've got. Harmonizer, really, in minutes, can just go through, and it'll identify any paragraph, or sentence, or what have you that is reused across the whole content set. It'll say "Hey, this particular paragraph is in 47 different documents," or it'll also say "This particular paragraph is in 48 other documents," or at least it's pretty close. It's within a few words. So we do this idea of close matches and – anyway, you can contact us about that, but I think that's another great way to really prioritize your content for that.
Marianne Calilhanna
Well, thank you, both. We are approaching the top of the hour. I want to thank everyone who's taken time out of their morning, afternoon, wherever you are in the world, to attend today's webinar. I'd like to remind everyone that the DCL Learning Series comprises webinars such as this, a monthly newsletter, and our blog. You can access many other webinars related to content structure, XML standards, life sciences content, AI, and more from the On-Demand Webinar section of our website, dataconversionlaboratory.com. We do hope to see you at future webinars. Please reach out to David and Regina. These are two experts available as a resource to you, and we hope everyone has a great day. Thanks so much for joining.
David Turner
Thanks, all.
Marianne Calilhanna
This concludes today's broadcast.