Zohra: 0:01

Hello folks, welcome to another season of Inside Techcom with Zahara Madabana. In season six, we unpack how generative AI works and what it means for your techcom workflow. From core concepts to practical use, we're gonna go under the hood. It's time to adapt, create, and thrive with AI. Let's dive in. Hello listeners. Welcome to another episode of Inside Techcom with Zora Mutabana. As we continue to explore and go under the hood and explore behind the hype and try to actually look at the invisible architecture that makes AI actually useful in content operations. And if there's one component that powers everything from smarter search to better personalization and content reuse, it's metadata. So that's exactly what we're going to dig into meta tagging and metadata strategies with Regina Lynn Preciado, who is the senior director of content strategy solutions at content rules, and Max Wischer, the director of technology. Together, they will bring the strategy and the tech sides of structured content and metadata into focus for us, especially for uh large complex organizations that are navigating this AI digital transformation. So let's dive in. Regina and Max, welcome to my episode. Thank you. We're happy to be here. Thank you. Thank you. Regina, would you like to go ahead and introduce yourself?

Regina: 1:38

Sure. As you said, I'm Regina Lynn Presciato, and I have worked with content rules since 2001 in various capacities. My background is in, I would say, components-based authoring, structured content, and metadata for sure. And I have spent decades in this space. Awesome.

Zohra: 2:00

And uh Max, please introduce yourself as well for us.

Max: 2:03

I'm Max Swisher, Director of Technology with Content Rules. I've been a technical writer and system administrator for quite a portion of my life. And these days I help customers in a wide variety of industries implement structured authoring and structured data solutions.

Zohra: 2:19

That's awesome. And these are things that as a technical writer that I do every single day. And that's what I'm here to unpack and actually personally learn from what tips I can take away, and hopefully our audience as well. So both of you, welcome again. And I want to dive right into before we kind of get into the mechanics, I want to kind of zoom out a little bit. And I want to how would you define AI readiness when it comes to content? And where does metadata fit into that big picture? Just to the level set. I will leave it to either of you, whoever would wants to take that question.

Regina: 2:56

I'll go ahead and take that. I think when it comes to AI readiness and content, the more structured the better. And having meta tagging at any level provides additional context and structure for whatever AI solution you're looking to implement. There's a quite wide range of how content can be stored, represented, and used, especially depending on how your organization started and what role information plays, both internally and for your customers. And for some companies, having boxes of paperwork from 2001 laying around is still a perfectly usable way of getting most of their daily operations done. And you can range from that to companies that have advanced all of their legacy systems onto data fabrics that enable sort of harmonized information to flow across parts of the business. And depending on where you are on that scale, there's always going to be some AI solution that purports to be able to do everything and more. And understanding where your company lies sort of on that overall AI maturity model, just in terms of the content readiness, not even in terms of the implementation or use of AI, is just really important to be able to gauge expectations and truly have successful AI projects internally.

Zohra: 4:20

As I was listening to you, you mentioned data fabrics, content readiness. And I think it would be great if we could explain that. But I'm going to give this question to Regina. Regina, you mentioned components-based authoring. And again, Max, you mentioned structured authoring. So I kind of want to tie all this in and where does metadata slash meta targing fit into this ecosystem?

Regina: 4:43

I think technical writers who have worked in either topic-based systems or component-based or structured systems, many of us think of these nuances, but in general, we get used to the idea of segmenting our content out into what I think of as one question and one answer. Even when it's not written as a question and answer, we really think about this unit of content answers this question or provides this information. And under the hood, that is often tagged with some kind of meaningful markups, such as DIDA, XML, or other schemas as well, that help humans and machines understand what the content is. And that fits into that overall bucket of meta tagging because it's a meaningful set of tags for humans and machines. And then when we go to the next level and think of metadata, which many of us have done where we tag a component, you know, this is for a system admin or an end user, or this is for California and that is for Texas or something, you know, it's very general tagging. We're often doing it to help ourselves find the information, define relationships, use search filters, et cetera. So when you take that basic, each unit of content, each component of content provides one piece of information. And within the content, it has some markup that describe, you know, I don't know, this is a step in a task, or this is a definition, or, you know, something. And you have the component tagged with things like audience and purpose and relationship. You get a very rich set of information about the content. And that is what our machines can leverage to connect sort of the human intent when you come to an AI bot and you're asking it for something to the content that has been identified by a knowledgeable professional to say this content is this and it's for this purpose, et cetera. So I think I answered the question.

Zohra: 6:46

Let me know if I perfect. Yes, absolutely. You grounded this discussion for me. I think you brought it all together. The one clarification that I do have before we again go deeper into this discussion is you mentioned data and XML. If we are not using, let's say, the data framework, this discussion would still apply, right? Absolutely.

Regina: 7:09

And I think what I see in my various customers that I work with, particularly in life sciences, is we have this whole spectrum of let's use AI to create more Word documents faster, all the way up through let's rethink everything as a digital first structured database of information with meta tagging so that we can leverage, you know, AI and other types of automation, better search, better assembly of content, et cetera. So yes, it it is just handy to kind of use data XML because when we talk with technical communicators, that seems to be the structure most people are more familiar with, even if they're not using it day to day.

Zohra: 7:52

Perfect. Max, coming back to you, you mentioned data fabrics. I'm curious, what does it mean in your world? And honestly, I'm hearing this for the first time, actually.

Max: 8:01

Yeah, I think it's sort of a new term that no pun intended is a bit of a blanket term, but I think it refers to the practice of really fostering automations and integrations and interoperability between systems and harmonizing data. I just want to throw out that metadata is really just data about data. And for us as content people, we think of data about the things that we're writing because that is the data that we're responsible for. But in other parts of the organization, whether that's sales or marketing or research and development, there may be other types of data that could also be tagged and be related to the same concepts that your content is related to. And when you develop actual systems that are connected and harmonize these concepts across different systems, whether it's a data system for your content and your documents that are being published, or a data system about the various resources in your company, or perhaps a domain knowledge base about anything else having to do with what your company produces. Just having that fabric, as they call it, of information that's harmonized. Different machines, even though they're built by different companies, being able to sort of speak the same language. I think that's what I think of as a data fabric in my space.

Zohra: 9:22

That's so cool. I'm almost thinking of something coming together to be woven and stitched together. And I'm showing that with my hands, audience. I'm I I I speak with my hands. But that's a beautiful metaphor, actually, to me. It's uh it's beautiful to think about how content is working at scale. Um across um uh teams, I would say.

Max: 9:43

Right. And it really is just one type of data that's produced in an organization, and being able to facilitate insights and relationships across different databases can really be powerful just for the business, even aside from the content production house itself.

Zohra: 10:01

That's true. So I'm gonna kind of flip my script here just a little bit and challenge myself. That's the goal. We've talked about what metadata is, what meta tagging is, and and just the whole ecosystem, just to level set what our context and our scope is gonna be. But before I dive in, I want to understand what are some of the challenges that you have run into before we dive into the solutions itself. Like companies are talking about this, writers are talking about this. In my own professional world, we are dealing, we are asking these necessary questions as we kind of, okay, all the hype is done. What do we do with our content? And as we look at that, it is it's not fun. It is messy. So I want you to kind of draw some picture around that and just help us.

Regina: 10:50

I think we often at content rules we talk about, you know, we want to identify sort of the the root of the challenge. Is it people? Is it process? Is it content? Is it tech? And, you know, as technical communicators, probably everyone listening understands that a lot of business folks don't think about content as a thing. They understand process, they're trying to optimize process, they understand technology, or you know, trying to use technology, and then the content gets a bit lost. Um, so I would say for challenges, most of what I run into is the people. Because, you know, most of us don't really have a deep grasp of what AI is doing, and you can oversimplify and say, well, it's just predictive text, but really fast. Or you can get into the full-on details of what's happening. But the challenges have been more about learning and getting through the fear, learning how to use these tools, understanding the the connection between well-structured, well-tagged content, when of course the first demo anyone sees is just let's give it some PDFs and let it make a summary, or let me help you write your email, you know, which is technical writers, we often that's the one thing we don't need help with. I think challenges will come up throughout, both technical and people. I think the key for people to understand is there are challenges with persuading and inspiring all content creators to tag their content. Everyone needs to align to a metadata standard to tag in a consistent way. I mean, there might be different ones throughout an enterprise, but you know, you it kind of has to be, there has to be little equivalencies and synonym lists and translations in between and so on. Um, but it can't be each individual writing a bunch of hashtags. The machines do a better job when the content structures and tags are consistent and predictable.

Zohra: 12:48

In my experience, content is rarely similar across functions, right? And you mentioned that the biggest challenge is the people.

Regina: 12:58

The biggest challenge has been people, in part because of where I sit in the more strategic yes, everyone has access now to Google Gemini and Microsoft Copilot and some other kind of internal to your organization. It just comes Acrobat, Adobe Acrobat has a bunch of AIs it wants you to use now. So, and a lot of people in our industry have experimented with just the free chat GPT and and meta that are open. So, how do you bring that all together into truly helping people be faster, better, more quality, accurate in technical communications has been, you know, there's a lot of conversation and there's a lot of pilot and proof of concept that was very exciting. And how you take that into operations can be very challenging.

Zohra: 13:47

Thanks for taking that question, Regina. What I would maybe segue into is now let's kind of bring this together. Okay, we run into this challenge. You have run into this challenge, and this is a challenge that's going to exist, and we'll have to find creative ways to overcome. But what are some insights? I'm thinking about a company where there is this communication breakdown. Rarely are different departments talking to each other, especially I'm thinking about support, marketing, and technical communications. Although my focus is technical communications, as you you rightly mentioned, content flows through an entire organization and it has to be seamless. How do we break down these silos at a high level before we get into the mechanics? I know I want to get into the mechanics, but I want to address these elephants sitting in the room.

Regina: 14:35

Yes. And I think we've seen a lot of talk in recent years about bridging silos, because we all know that organizations are too big, it is not practical for everyone to be in the same silo. And this is where having a standard, and I like to talk about having a standard in two ways. One is kind of your enterprise metadata strategy, which is common categories or facets of metadata with common labels that everybody can use. And with AI, we start taking that one step further into ontologies and knowledge graphs, and so it's relationships across these things. But there's a common standard for the whole enterprise. And that way, no matter where you sit, TechCom support, learning, development, internal training, commercials, marketing, wherever, you have a one-stop common place to go to that is hopefully governed by a central function. And then every use case, every type of content might have additional metadata tagging. And a very simple example of this is if you have metadata for your component content that's mostly text, you know that you will have additional sets of tags to put on media assets like images and videos and things. So if you take that idea and you say, okay, here's our common enterprise metadata tags that people will use for everything. And then technical communications might have a separate set or an additional set of tags that is really the technical pieces. And learning and education will have its own set of tags, maybe related to e-learning and things that aren't really relevant to other types of content. So that's some examples of how you bridge the silos where start with a common, but you also support each individual team. And I always come back to not just the people, but the content types with the relevant metadata.

Zohra: 16:33

Again, you simplified it for me. In my head, as you were explaining this, I was sort of drawing a wen diagram. But before I rephrase what you shared with me, Max, I want you to share any perspective you may have.

Max: 16:46

Yeah, I think as Regina describes, there's always across an organization significant overlap in the concepts that different groups are working in. There's also significant pieces of information or parts of the knowledge domain that are so specialized to one given division of the organization that maybe having all of those options in your list of values that you can pick from is not necessarily, you know, not every tag should be available to everyone to use all the time. But when folks are using tags that are the same across different systems and across different groups, having a unifying layer that really knits together, using that pun again, all of the different sources of information so that they can be leveraged in tandem moving forward is really important organizationally and also just in terms of making data usable for AI across silos and not just within one.

Zohra: 17:42

I'm so happy that you brought that data fabric metaphor back into the picture with a great example. My takeaway is we are thinking about content governance here and how we start looking at the relationship between content that is siloed and thinking about ontologies and graphs, which we are obviously not going to dive into. I think that's a very deep conversation, even though we are going to dive deep. So this is where I'm going to dive deep now and jump into this question. So, how does effective meta tagging serve as a key step in transforming this messy content that we've been talking about to make it AI-ready and user-friendly?

Max: 18:21

I'll take a try at that one. From my perspective, when it comes to at least large language models and what we think of as the modern incarnation of AI for content, either on the author side or on the user side, leveraging it to generate information about what's already been authored so that they can either solve a problem or create better information in the future. In order to effectively do that, the best solutions that we've seen are retrieval augmented generation systems, also known as RAG. And folks who have been shopping for AI solutions have probably run across a host of tools that have some form of RAG. And just to try and explain it in a very quick nutshell, this is going to be a gross oversimplification for the sake of brevity. Basically, instead of taking a huge corpus of information and training a language model on that corpus, instead, every time a query is made, a rag system performs some sort of search, usually using a method that does not involve an LLM, and picks certain assets from a given set of data to provide to an LLM, hopefully that was trained with some overall domain knowledge, though perhaps not the specific knowledge of the content you're querying. That way it's able to leverage the domain knowledge and the specificity of information that a search warrants in order to present results that are actually traceable to content. Because the other approach that is not RAG, as which is the term for retrieval augmented generation, approaches that do not rely on RAG tend to have much less of a clear path from the answer back to the source information. And depending on the risk level of the information being presented, that could be a really big difference. So when we think about how a RAG system works and the way that the first step, the actual filtering of information out of the database to provide to the LLM, happens independently of the language generation itself. That is sort of the first stop for where granular information that has solid metadata on it can allow for the proper information to get delivered to that language model in the first place. That's sort of step one in terms of how does tagging actually make a difference. And then I think step two, depending on the AI system, if you have this sort of fabric that I talked about, you have some sort of computer readable representation of how the different concepts at your company relate to each other. That will improve even more so the context that the language model itself has in order to interpret the information in the prompt. So that's an additional mechanism by which the additional data that the metadata around content pieces are providing to an AI system can enable it to produce truly solid and reliable results that can be used in real settings, other than just writing an email.

Zohra: 21:42

I'm going to try and paraphrase because you gave you unpacked a lot with your answer, and there are such good details in there. But one thing at a high level, what I'm taking away is meta tagging will help with the right systems in place, traceability of content, where is that content coming from? It is going to provide better context because if you have good meta tagging practices, that would translate to better findability of content. And then by the virtue of that practice, you're you're creating trust with your content for users.

Max: 22:13

Would that be accurate? I think so. In some ways, the tagging is both not the easiest step, not the fastest step. It's usually a very messy step as well, because coming up with a set of tags to use and truly understanding across an organization what they mean and then actually applying them to data, even if you have help of a computer, it can still really be a painstaking task. And as a result, usually that's the one that gets left out, especially because generally speaking, it's not in software companies' interest to tell you that making their tool work is hard. So as a result, they will tell you that without necessarily doing all of that work, they'll still be able to produce great results. And typically the meta tagging is something that's just sort of left out in terms of process. But it's a very important ingredient when you go to create and use AI systems that leverage that information. So it is both. You do have to have the information there to be leveraged, but just putting the information on the content isn't going to guarantee that whatever search engine you implemented today or 10 years ago or 15 years ago is going to be able to leverage that information. So it's both a people thing, having the awareness and the leadership to make tagging information a priority and see sort of the long-term value, even though it might not seem like it's super valuable immediately when you're checking the boxes in the content system, all the way through to having a system that wisely uses the information that's available and can harmonize that information with the vocabulary that you're using across different teams. And I think it's just important to keep in mind that it's both systems and people and actual data. And without any one of these three things, you're not going to have success.

Zohra: 24:10

I would agree on that. And I've heard of these challenges as I've spoken to other people as well. Regina, would you like to add anything to what Max? Yeah.

Regina: 24:20

I'd like to say that everything content creators should have been doing all along to make content more findable and usable for humans. You really have to do it now because the AI access to the content is going to expose the gaps and the inaccuracies. And I have a talk about governance as well. Everything we should have been doing to govern and have common workflows and lifestyles, we need to do now because AI will expose the gaps. The good news is about tagging things. One, you mostly have to do it the one time when the content is created. Most companies right now have a big backlog of untagged or inconsistently tagged content. But once it's tagged, the tags don't actually change all that often on a piece of information. And it's a good idea to review your metadata whenever you update a piece of information and you know, maybe new personas have been added and you need to add something. But so once it's tagged, two, if you cannot figure out what tags to apply to a component of content, that tells you your component is either too broad or doesn't really align with the important concepts of your company, or in rare cases, it's exposing that, oh hey, our metadata needs to be extended because we have this new type of content. So as an author and a creator, that metadata is really important in helping you. Do you have the right focus? Are you creating the right content? If there isn't a tag for it, do we even need to create this content or do we need to update our tags? It's part of authoring and making sure every piece of content that is created moves the business forward. And you're not creating it just because, well, our template from 10 years ago says we have to put this in, and so I should put it in.

Zohra: 26:18

This is really pausing me to think where I could apply what you said, but would you be able to illustrate this with an example, uh, Regina or Max, either one of you? If I have a unit of content and I want to add metadata to it, what are the questions I would ask? Of course, we talked about the persona, the relationship with other content, any attributes I would think of as metadata as well. But can you give me one specific example just to kind of deconstruct this and level set me?

Regina: 26:50

Yeah. So one of my recent projects is or was about medical information and specifically about cancer information, science-based, but written for lay people. So not quite the technical writing like a user manual and online help. But the cool thing is it's public-facing information that people can go to on the web to look up information about cancer. Now we know a lot of times a person looking up information about cancer is upset. They're recently diagnosed, their family members diagnosed, they're three years into a cancer journey. It is not your happiest moment, let's say. So this information needs to be findable and reliable. So those in working on componentizing this information, the metadata became the most important part of our componentization. What question is this unit of information answering? Is it answering something anatomical? Is it answering about treatments, prognosis, procedures that a person might have to go through? Are we talking to a patient or a caregiver or healthcare professional? So by tagging that what used to be long-form articles that were extremely well written, so important and very good content, chunking it up and tagging it to make it more findable more quickly to people who are in distress, served the audience much better. And now that people are starting to access things by asking AI or having the AI overview just thrust at you through Google when what you were really looking for is, well, where is my reliable source? You know, this is making a tremendous difference in getting the information to the people quickly and reliably, knowing that they're not in a state where they want to spend 30 minutes trying to figure out what their search results are.

Zohra: 28:45

Thank you for that example. It really kind of creates this sense of empathy as well as I hear about this, because I'm working purely in tech. And there are these scenarios that I would not commonly think about. So it's such a great example, Regina. Thanks for sharing that.

Regina: 29:00

And it works in tech too, because people don't come to your documentation when they're on a happy path just to find out about your tool. I mean, maybe a little bit. People come because they're trying to get something done. They can't do it either because they don't know how or it's not working, and they've just got to get the answer and move on. And we always say the the best technical documentation is invisible. People forget they even looked at it. They just got their task done. And I think the AI is really exposing redundancy and gaps and missing pieces and old information because the AI tools don't quote, quote, you know, in air quotes here, know what is most relevant, what's most current. The better tagged your content is, the better the AI can support your user.

Zohra: 29:46

Perfect. Very true, very true. Max, do you want to add anything?

Regina: 29:51

As another example. And are we just sort of talking about like examples of use cases for metadata within AI or not? Not necessarily within AI. It could be either. I think it really depends so much on sort of where your starting point is. I think for businesses where the current landscape is folks are creating documents in Microsoft SharePoint and they're just creating Word documents. And that's their current, the entire infrastructure that they have for writing information. At that level, there's a lot you can do in SharePoint with terms and organizing the information and making a SharePoint database that is very usable and also can provide information to a front end that, depending on certain criteria, like what product you already own based on the account you're logged into.

Zohra: 30:45

If you have some metadata and then you build upon it, then there are better outcomes. I think that's my takeaway.

Regina: 30:51

I was thinking that I've heard some vendors, AI software vendors, who really want to be welcoming and easy for their customers. They have sort of the metadata is something that happens and they they handle it sort of technically. And there are some tools that can automate some tagging that can get smarter about extrapolating from the content itself what the tag should be that can get you part way. But I think as content experts, as content professionals, we are the best, most knowledgeable source of the intent of the content, the type, the purpose. And again, if we can't figure out how to tag it, we probably haven't refined the content enough. So there's real value in having that indexer brain or the librarian brain, or if you don't feel yet, if you're someone who's, you know, don't feel yet like you know what you would tag, find someone on the TechCom team because somebody, there is somebody who is hyper focused on that too. So we can all help each other. Because again, everything we're doing to better identify and focus this content to be better processed by AI or even just old-timey automation from five years ago makes the content more understandable for humans, more reusable, more of an asset to the business. And again, you will also see for yourselves where your content planning for next year needs to go. Because one thing a lot of companies are using Gen AI for is to create a bunch of derivatives. In pharmaceutical companies, you have a very scientific document, the clinical trial protocol, that explains what's going to happen in this clinical trial, who needs to be in the trial, what the drug is all about, et cetera. Well, that's very that's not layperson friendly. So there's some experiments with using AI to generate a layperson friendly version of this to help with recruiting patients to participate in a clinical trial. So that if, you know, we have a someone who has a condition, I don't know, let me pick something, uh, diabetes, and who might be participating in a clinical trial for one of the GLP1 medicines or something, they need to see a layperson-friendly document and maybe something that's even more personalized, like, hey, you live in an urban area or you live in a rural area, or you're in a particular region or country. And the AI systems can get good at taking the original scientific document or documentation and sort of creating these derivatives. And where the metadata comes in, of course, is to be like, yes, use this, don't use this. This is internal information, this is external facing, this is important for patients in this area, this is important for patients in a different area, et cetera, et cetera. So again, as content creators, we're the ones who know, who have the subject matter expertise to mark up this content for the various types of machines, now including AI, to provide real business value and serve the customer as well. And I know it's scary because a lot of executives are jumping right to how many people can we fire because the AI can do it. And I think the the short-sightedness in that is the content that AI is creating is all derivative. And if it can't find something, it might make something up and therefore it's still derivative, but it's inaccurate. And, you know, the value of technical communicators and other content creators in creating really good source information for these machines to then automatically deliver or or derive or translate. I don't even literally mean translate into other languages, but translate into for various audiences, highlight something for one audience and highlight a different thing for a different audience. It's just super important to have your source be clear and accurate and tagged. There, I'll get off my soapbox now.

Zohra: 34:53

No, that was that was beautifully, I would say, layered for me. Two important things that I want to come back to content is an asset and business value. When you start positioning your role in that context, I think as writers, we are all definitely worried about are we gonna have a job or not? But then you just brought with this example brought to surface how we need to be thinking about this and then how AI can quote unquote air quotes hallucinate if things upstream are not done right, and what those could be.

Regina: 35:29

We talk a lot internally at content rules about knowledge capture, getting that information into some kind of digital format is important, but just capturing it, you know, I could blather all day into some kind of transcript and now it's captured. I could then hand it to an AI and say, hey, do your best to chunk this out and identify the themes and put metadata on it. If I did that and then I said, okay, here's the content rules sort of repository of information about metadata, it's not going to be reliable in the way that it would be if it is then reviewed by our metadata experts and it is tagged by an expert indexer who might be my, might be me, might be one of my expert indexer consultants. But at that point, that knowledge becomes a business asset. It's up to that point, it's like probably close enough for internal use, maybe, but it is not something we would expect our customers to rely on.

Zohra: 36:27

Yeah, with all the examples that you both have provided, I think it's become even more important for us why we should consider having you humans in the loop and experts in the loop, and why content creators are the best to be at that driver's seat.

unknown: 36:43

Yeah.

Max: 36:45

An effective implementation of uh taxonomy and metadata across an organization really has to be a common expression of the knowledge domain of the business. And every business has a unique knowledge domain. I mean, there might be interconnectedness. You might have a business that works in the energy space, and there are domains of knowledge about how the world responds to certain types of pollutants or other things that are tied into your knowledge. But when it comes to specifically what your company does and what your company knows across all of its uh departments, that's when having not just tags, but information about the tags and a common language for what a given tag means. You cannot just let a computer loose, coming up with its idea of what something is about, and then expect the results that machines are generating to actually be relevant to the domain knowledge of your company. So it's really sort of imparting that human layer that isn't automatically embedded anywhere else in the system, but creating a very discrete and explicit expression of what it is that your company is about that can, I think, really provide the most business value. So as you said, human in the loop, having the experts, having the content creators, having the folks who are generating the information. And by generating, I don't mean generating with AI. I mean creating based on their understanding and their cooperation within the business is really key.

Zohra: 38:26

Yep. And again, thank you for driving home that point, Max. It this this discussion is so important as to why we need to continue with this human in the loop, human oversight as we integrate AI systems into our content workflows. Yeah. We have quite a few more questions, and I know we are coming up on top of the hour, but I do want, I think we've covered quite a few. One of the questions I had was what are the practical strategies for aligning cross-functional teams around consistent metadata standards? I think, Regina, you answered that question. So I'll unless you have more to add, I would like to skip maybe a few questions down the road. Do I have the heads up from you both on that? All right. I mean, we didn't talk about specific tools, but there's this one question. Which principles or approaches are most important for creating metadata schemas that support both AI needs and human goals? Again, we've touched upon this, but I would like to specifically have you both answer that for me.

Max: 39:23

I'll go first. When I think about the modern day expression of data tagging and metadata tagging, I think that knowledge graphs and sort of rich taxonomy management that goes beyond just a list of words is really where sort of the potential lies these days. And there's nothing that prevents, as I said before, not having every tag available to everyone at all times is crucial because there's going to be so many different areas of information in your business that that's just not realistic or practical. But having a common understanding of how different tags relate to each other beyond perhaps just broader and narrower within a hierarchical list, perhaps even having different words that mean the same thing and having those defined in a system that understands that these two words mean the same thing. And depending on the organizational needs, you can allow two different departments to call to, to call the same thing two different words, if you really have to, when you have a system like this. And as long as your domain knowledge model makes those synonyms and presents to the systems that are accessing the data the fact that when you look at these two separate assets and they have these two separate labels, they're actually tagged with the same common concept, just with different labels. This is a concept that also applies across translations, because if you have a global business, you need to be able to relate all of the different translations of a given concept together and have a common understanding across the business of which words mean the same thing across languages and which words don't mean the same thing across languages. And having that information in a place that's really accessible to everyone in the business so that it can be a common understanding that underpins all the information that's created. I think that sort of next level of relationships between and labels for metadata and taxonomy facets is really where it's at.

Zohra: 41:32

Awesome. Regina, would you like to add something there? I think that covered everything I would have said. I think so. I think that's also too good. That was that was great. That was really great. I'm going to jump into my next question. What tools or platforms do you recommend, other than coming to content rules for advice? I would I would recommend that as the first place to start. But if we had to consider any tools that you can recommend for somebody trying to explore Yeah, that is funny because we do help companies identify their actual requirements, not just the requirements they thought they had, and evaluate tools and provide recommendations.

Regina: 42:12

And I think I can't name tools on this. We we have to kind of stick more with tool types because we recommend component-based authoring. And it often that goes for technical content that often means a DITA system, but it doesn't have to be. There are other components and topic-based systems that depending on the scale and the size and all of that would work. I always like to have a uh what I always called a content optimization tool that plugs in that helps with terminology, can notice that, oh, you use this old term, we suggest this other term with voice and tone. There are several of those out there as well. And Max just answered about some of the more you know taxonomy manager terminology management. So I think that is the core. The core is some kind of component structured repository content optimization. You will need a localization manager tool as well, because we need to be providing proved reliable content in many languages, and there's no excuse. The excuse for a long time has been cost, and that's not an excuse.

Zohra: 43:24

Regina, I love how smoothly you answered this. You just sailed through it. I need to learn a thing or two from you on how to be diplomatic but still answer that question. You gave some, I think that was perfect. Authoring tools, optimization tools, localization managers. I mean, you kind of gave the lay of the land, not specific to, and that's what we are looking for. So thank you for taking that question and answering it in this smooth fashion.

Max: 43:51

Yeah. And again, as Regina said, we we spend a lot of our time working with customers simply to evaluate their actual requirements and figure out if a new tool is really the right way to go, or perhaps provide external validation for a team that feels pretty confident but wants to have a little bit more teeth in their proposal to upper management. And sometimes we discover that the problem has nothing to do with the tool, and there's a lot that could be done without getting a new tool or incurring additional cost. And sometimes we're able to make matches that are made in heaven and produce really amazing results by implementing new technologies. So we are definitely folks who focus around technology is one of the tools in our toolbox, but we definitely are problem-focused and solutions focused, and tools are just one of the ways that we go about doing that.

Zohra: 44:50

My audience, if you're looking for a marriage made in heaven, please hit up Regina and Max. How about that? I gave you both a big plug because I want a marriage made in heaven. That's a good one, Max. Moving on, we have just a couple more questions. I think probably we did talk about this. How can teams integrate meta tagging into existing authoring and review workflows without adding extra burden? Max, your answer right now was about it's not just the tools. There are so many other things to be looking at. But is there anything else either of you would like to add?

Regina: 45:23

Metadata is not a burden. That's what I would add. Metadata is part of authoring. And if you don't tag it, who's gonna tag it? Some system that that some other department trained? No. So metadata is part of authoring, is a content rules philosophy.

Max: 45:41

There are tools that can help, but having a human to make sure that they are accurate and doing so consistently and treating it as part of your process is really important.

Zohra: 45:50

I think so too. And that's something that I'm honestly paying more attention to. Because content, we inherit content, right? Somebody else created that content. I inherit it. So I have to revisit and do that audit. And I think at my company, we have this really good practice where at the end of the year we sort of go and revisit and do some sort of a content audit. And it gives me some other time to look at, okay, what new content did I create? Did I add meta tags to that? But let me also look at the old content. Is it relevant? Yeah, especially as we start using AI tools in our content workflow. How do you balance machine readable metadata with human-friendly content management? Like I'm thinking with this question, can I overdo with the meta tagging? But um yes, please go for it.

Max: 46:40

So the world of taxonomists is just an entire profession and art and philosophy of trying to figure out how big of a box is the right size box for a given idea. Because you can always go broader and you can always go narrower. And the sweet spot is always somewhere in the middle. And large organizations are going to have very large domain models, probably, but there always comes a point where it's not useful anymore. So I think it really depends on the use case. For some applications, you can just select the country. And because it only changes from country to country, that's fine. Now, if everything is tagged with a given country that's in, let's say that every object that's tagged with countries is always tagged with the same countries that are in a continent, consistently, always, 100% of the time. That's just sort of a an obvious example where, well, we don't need all of these tags. We're using duplicate tags to mean the same thing. What are we actually trying to explain with these tags? Are we trying to explain the continent? Are we trying to explain certain aspects of climate that are common across countries? Sort of doing an assessment of is the granularity enough for our use case and enough for it to be useful, while also not being so much that everything gets a huge number of tags that aren't actually meaningful. So that's sort of the in-between in terms of having useful information, but not having too much information.

Zohra: 48:18

Yeah.

Regina: 48:19

This is where you can learn, learn as you go. If you do too broad, you'll find you're not getting the results that you want. If you do too narrow, your authors will rebel, probably. Starts to interfere. So there is a little sweet spot, as Max said. And then not to feel like the first one, the first taxonomy metadata knowledge graph you come up with is you're locked in forever, because that's not true. And it will change because your business will evolve, your tags will evolve, what we call things will evolve. So I hope that's reassuring to people. Get as close as you can, and then let the results guide your iterative. I love Zora that you said your team does a kind of a content audit at the end of the year.

unknown: 49:02

Yeah.

Regina: 49:02

And that's that's the governance piece that requires, you know, that again, AI will expose if you don't have governance. So it's just a good practice.

Zohra: 49:13

Yeah, I completely agree. And I've been at other places where legacy content is never looked at. And when I say legacy content, it's been around for 40 years. Yeah. Um, and it's been written in different voices, different tones, it's lost relevance, but it's still there. And nobody wants to prioritize retiring that content. So I definitely appreciate when content audit happens where you get a chance to for these exact reasons that we've talked about business value and content, where you're trying to show content as an asset.

Regina: 49:43

When you have some non-personal criteria that a group can agree on, like, is this increasing revenue? Is this decreasing costs? Does this save us a headache? Is this regulatory required that we that we have this forever? You know, if you're working in pharma, yes, everything you've ever written about a drug, you have to keep forever. So, or you know, sometimes only 50 years. So you have, but then you can easily evaluate more easily than the gut feel of, oh, but we work so hard, or what if we need this someday?

unknown: 50:16

Yeah.

Regina: 50:16

But I'm sorry, I interrupted you.

Zohra: 50:18

No, no, no, Regina, valid points, extremely valid points. I'm just trying to think in terms of, you know, when I think about a use case, the examples that you've provided, the questions and the insights, it makes it easier for me. So when I'm thinking about meta tagging, do I need to go narrow? Do I need to go deep? How does that tie to the business value? Those inform how I can address the lack or the excess of meta tagging in my content. So I really like how we've wrapped this up.

Regina: 50:47

I just want to say thank you for having us. And that when we met you at the um, what was it, the convex conference, I think it was a structured authoring content. We were both super impressed, Zora, with your knowledge and your experience and so excited to get invited to be on your podcast. So this has been awesome to talk about. We could go on. I know we both have other things we have, all three of us have things we have to get to today. But yeah, just thank you. And this was a great discussion. I got things to think about further as well.

Zohra: 51:18

Thank you so much. I really, really appreciate that from the bottom of my heart. Um, thank you.

Max: 51:23

Uh yeah, want to echo what Regina said. Thanks for having us on. I want to just throw out at the end that even if you don't have a huge business, metadata can still provide value for you. And even just thinking about small lists across a small business of words and what they mean and putting them in a place where you all agree that they live and starting to just organize whatever files you might have laying around, or maybe start using those words on your slides when you're starting to talk about certain ideas, can really produce a lot of different value internally that can also set you up for success in the future. So there's no point that is too small to start with, and the dividends will only pay off more and more into the future.

Zohra: 52:11

And on that note, Max, this is a great wrap. Thank you both for your time and for educating me. I love these personal conversations and I've learned so much. Great insights. Thank you, and have a lovely day, Regina and Max. Thank you. Listen to InsightTechcom on your favorite app and follow me on LinkedIn or visit me at www.inside techcom.show. Catch you soon on another episode. Thank you for listening. Bye bye.