Shaping the Future of Knowledge Management: Collaborative Innovation in Intelligent Content, Taxonomies, and AI

 Alvin Reyes, of RWS, Harald Statlbauer, of NINEFEB, Mark Gross, of Data Conversion Laboratory, and Lance Cummings, of UNC Wilmington, were part of a Component Content Alliance-sponsored panel discussion to wrap up the conference. 

"If you have one dollar to invest in knowledge management, put one cent into information management and 99 cents into human interaction." - Larry Prusak

Preconditions for AI-driven solutions include organizational readiness, standardized metadata, and appreciation and application of structured knowledge and structured content.  

What are the key starting points for orgs to unlock the potential of their knowledge assets with taxonomies, intelligent content, and AI? 

RAG and AI are ways of getting info out at a faster and better rate than in the last, but basics have not changed all that much. People have come back to thinking that content has to be good content and has to be structured content. GenAI is output. Combining unstructured and structured methods to get advantage of structure. People are writing structured content, throwing away the structure publishing to PDF, and then using the PDF to feed LLMs. 

What roles and skills are essential to AI-driven knowledge management?

Work across teams, but writers will be at the table. Bridging gaps between technical documentation and learning and training.  Working across teams will require diplomacy skills. 

How can orgs create smarter workflows by aligning AI, taxonomies, and content strategies?

Start with terminology. Out of that, creating a taxonomy.  Workflows are driven by end results. Work backwards to support that. Content is messy, and most of the content you're going to be using is not content you have control over. Have to figure out how to add structure. This is where taxonomies are important. 

What's the relationship between structured content, structured documentation, structured knowledge, and AI?

In any complex system, both structured and unstructured information. Knowledge is mirrored in different types of media. Reclassification of content helps make it accessible. RAG is the approach for your content, not a random LLM. Context needs to know where the content is coming from. 

 

The Journey to "Docs as Code" and Back: Lessons Learned from Three Documentation Migrations

 Mike McGinnis, of Tridium, shared Tridium's journey of multiple docs migrations. The size of the team and the type of the people are factors in how you do migrations and whether to do migrations. 

Environment is agile, mostly software, some hardware, publishing to PDF, in-product help, and a document portal. 

The first migration was from FrameMaker to Windchill (Arbortext). Had too many files, DITA looked promising, wanted to avoid vendor lock-in. The migration took 2 years, Arbortext required separate training and tooling. 

Migration #2 was Windchill to Git, because the Windchill environment was being lost. Had short notice, and saw docs as code as an option. Had experience with Git and knew could set up a repo. Export not as easy as thought, had to export one book map at a time, and files were put in one folder. Had to refactor all references, mostly done with scripting. Writers needed to install and learn Git and Sourcetree. 

Docs as code led to merge conflicts and confusion because of no training, and it was easy to break things. We also had no dedicated support, which was a problem. 

Migration #3 was docs as code to CCMS. Publishing effort was high, legacy tools were expiring, writers were unhappy, the vision was unclear. Redefining the requirements need to include "easy." Easy system architecture, admin controls, operational controls, authoring, publishing, and reviewing. 

Lead people, not just tech. Create shared vision, and empower others. Adapt and reflect.

Structure Through Delivery: The New Gateway to DITA Adoption

 Frank Miller, of Ryffine, started by talking about DITA adoption over the years. Tried many things to get people to adopt structure and good content strategy. But there is resistance to change. Projects take too long and stall before showing value. Silos prevent cohesive adoption. Orgs struggle to justify costs, especially when benefits are years away. 

Today is a perfect time, with the knowledge and experience we have, and all the tools we have. This includes the AI revolution, the evolution of content delivery platforms, and market demand, when customers now require intelligent, contextual content delivery. Structured content is fast becoming a business imperative. 

So why now? We have years of implementation wisdom. Modern CDPs enable adoption. AI drives structured content demand. And there are now proven pathways to success. 

In the traditional approach, there was heavy structure, in the front end doing all the IA work, resource drain, high upfront costs in time and training, timeline risk, a long time to demonstrable ROI, significant resistance across teams, and benefits only after implementation. 

New approach starts with delivery, in modern CDPs, and show immediate value, have a gradual evolution, replacing a "big bang" with progressive structure, with early wins and flexible timing, all of which lowers risk. Give leaders a shiny object that is not only shiny, but valuable. That buys time, support, and budget. 

Content value can only be truly measured at delivery. We need feedback fast. 

Modern platforms support multiple formats seamlessly and connect with existing tools and workflows. We can use CDPs as aggregation points for our content. 

Now, instead of having all content in a single repository, the dream, teams want their own repos, so having all the content in one CDP is a good compromise. 

You have to do the metadata/taxonomy to get the output right.

Content Conversion Hacks: Tips and Tricks for Success

 Adelheid Certik, of Mizuho OSI, recently made the transition from unstructured authoring to structured authoring. Here to talk about the content conversion. 

Prioritize the content to convert. Some immediate, some ongoing, some never. For example of the latter, older consent or content not a candidate for reuse. Also wanted to identify a subset of content for a pilot, one complex and relevant enough to develop a good content model.  

Do you want to outsource or do work in-house. This is a money vs. time equation. 

In pre-conversion, edit your content to align with your information model. Create styles to cover elements/attributes in your information model. Apply styles consistently and then break content into topics or chapters. 

For the actual conversion process, conversion tables are available as a plug-in in tools such as Oxygen and FrameMaker. These assign tags to styles. When going from XML to DITA, adjust the XML structure to align with the information model. If you have the in-house knowledge, can use XSLT.

After conversion, upload to CCMS, then create maps, add links to images, add conrefs, and create keys, keyrefs, and conkeyrefs. 

Do not skimp on editing your content to match your information model or identifying main areas of reuse. 

Bridging the Gap Between TechDoc and FieldService and Support

 Harald Stadlbauer, CEO of NINEFEB, noted that the session will deal with large, heavy equipment, and how to serve the service technician, often forgotten at a company, so they can service these types of products. 

iiRDS (intelligent information Request and Delivery Standard), an international standard for delivering technical content to service technicians. 

Gap between tech doc and field service, writing topics, usually published as PDFs. Service tech gets large PDF, too much info not tailored for specific situation. 

Out use case: intelligent content as a service. Content for the right context as a service when you need it. 

A service engineers wants context-specific information for problem resolution. 

iiRDS is an international standard that consists of a standardized vocabulary and relations between the vocabulary entries and classes.

The Docs Pipeline of the (Near) Future

 Manny Silva, of Skyflow, started by saying that he's going to talk about how the ways we create content is going to change. 

Today's pipeline consists of authoring, publishing, and collaboration tools. (C)CMS pipeline includes DITA tooling, WYSIWYG editors, PDF & HTML, and Sharepoint and Jira. Kind of looks like first get a request in Jira, author content in CMS, and then output. Docs-as-code gets a but more complicated. More tools, each one doing less. 

What about AI? Much digital ink has been spilled about how AI has upended everything. It's not going to replace writers. (But will change how we work.) Generative AI benefits from personal, targeted use. Not everything can or should be solved by AI. 

Challenges with current pipeline: content is created manually, time-consuming validation, and maintenance overhead. 

Three categories of new tools that will help automate routine tasks. 

Content generation. Tools that assist in creating documentation content. Rapidly produce first drafts, freeing us to focus on refinement. For creating novel technical content. 

But generation challenges include that there are a lot of generation tools already, with more coming out. 

Content testing, which embodies docs-as-tests as a strategy. Docs are testable assertions of product behavior. Validates information before publication. Not for validating syntax or style, which is what linters do, but that doesn't validate the actual content accuracy. 

Testing challenges include a learning curve (not traditional tech writer tools).

Content self-healing tools represent the cutting edge of documentation automation. They automatically identify and resolve common documentation issues with minimal intervention. 

Self-healing tools require a deep understanding of technical documentation content and why the structure is important. They don't understand IA and why we do the things we do. There are also security issues. How can you be sure that there will be a human in the loop?

Thyese advancements are not about replacing technical writers, but about empowering them. They allow tech writers to focus on high-value tasks and produce more accurate, trustworthy documentation.  You can maintain larger doc sets and adapt quickly to product changes. This elevates the role of tech comm.

Fighting Words: DITA and the Battle for Better Content

 Jake Campbell, of Scriptorium, assumed a basic (basic) knowledge of DITA to start his presentation, described a problem of a lot of content spread across several poor-quality PDFs with no bookmarks, no searchable text, and file size issues. 

Initial solution extracted data into spreadsheets. A newer approach was a modern PDF, but still issues. 

Core issues were irrelevant information, information accessibility, and resource usability. 

Solution was to collect into one PDF with navigation and clickable links. 

Used DITA-OT 3.6.1, oXygen, Antenna House Formatter, and Bitbucket. 

Process included selecting conversion target, build a content model, decide on a workflow, and then do the writing. 

Created a template with section elements, most with titles. Added Details and Tags sections. An otherprops added to the root element allows for filtering.

From Writers to Engineers: The Evolution of Product Content Engineering

 Beth Lemesany and Dawn Bunting, of ServiceNow, noted that we weren't just just supporting content, but engineering the systems behind it. Writers wear a lot of different hats. Important to have clarity in your role. 

Do you troubleshoots content issues, build content, use scripts? Do you consider how content scales, or work with engineers or tool admins? You might be an engineer already. 

When yo rebrand to an engineer, there is less context shifting, get to focus more on strategy and scale, getting pulled into the right conversations, and get built-in efficiencies. 

Product Content Engineering engineered conversion from wiki, built DITA-OT transforms, and supported writers with DITA training. 

Eventually moved from support to tools, become proactive instead of reactive. Started to think about scaling. Generated a classification map for search taxonomy. Created automated tests. Exposed to "engineering" without yet calling it that. 

Leared to think like engineers, started to deliver like them. Didn't just level up, but re-orged up. Learned to think in terms of projects, automation, scalability, and maintainability. Team had a PM, which helped in visibility. 

As scope grw, so did structure. Identifies type of work through mandatory ticketing. Split work into categories, such as support work, project work, and technical debt (later adding technical wealth), and leaned into agile, which allowed planning for capacity. That allowed respect for capacity, allowing us to say "no" more often, and allowing the ability to advocate for headcount. 

Best practices provide consistency and efficiency. Architecture diagram creates visibility and collaboration to help understand how the systems work, creating shared understanding and clarity and accountability. A change management framework protects the user experience, creates alignment across teams, and scales. Recurring meetings and ceremonies allow for structure and focus, helps catch and fix real problems, and makes work easier to sustain. Coding best practices keeps code maintainable, supports collaboration, and reduces risk. 

When we faced resistance and skepticism, we turned it around by building credibility through delivery.  

Scaling is harder than solving. That's why you give the work structure.

How We Live in the Shadows: Taxonomy in Large Wikis

 Jeffrey Scattini, Haley Helgesen, and Lindsay Bachman, of Crunchyroll, talked about a taxonomy in highly unstructured environments. 12 years without a standard IA and multiple knowledge repositories. Much info locked in people's heads. Acquired companies, none with doc people, folded in many doc environments. 

Current IA lift: 16,0000 pages across 70 spaces. Teams organized own content, and were attached to putting content where it made sense for them at the time. Widespead interest in findability, searchability, and relevance. 

Confluence: hero and villain. Inherently amorphous. 

Confluence operates as a "folksonomy." A collaborative classification system of applying user-created keywords to describe and categorize information. 

Designed and implemented a standardized hierarchical taxonomy, applied across all business units. It needs to serve a range of users. To be accessible to all teams, needed  to hear from all teams. 

Started with a card sorting exercise. This identified common patterns and rationalities, which helped created a taxonomic structure with primary, secondary, and tertiary labels. 

A parent-child label schema is the standard backend of a taxonomy. It reinforces the taxonomic language by using consistency and repetition. It fosters inter-document relationships. 

Implemented with a phased approach. Started with content analysis.  Then implementing based on what found in space reorganization. Content audit used Google Sheets with info about pages. Analysis of each page included whether it should be archived, retained, or reviewed, and its place in the taxonomy.

The Wizard of Docs: Finding the Magic Behind Taxonomy-Driven Documentation

 Scott Hudson, Sr. Mgr, Content Architecture, and Eliot Kimber, Sr. Staff Content Engineer, opened their presentation with a song about taxonomy using a very Wicked/Wizard of Oz theme. 

The land of documentation chaos (before taxonomy) customers were Dorothy, lost in a confusing landscape, no clear path, no consistency, really needed a Yellow Brick Road. 

How do we help Dorothy? How do we classify content in a way that works for both authors and customers? The goals: Better navigation, fully classified, and good governance. 

Classification has three major components: automated classification, authored classification, and governance.  The automated part is assisted by AI, but the authored is the human oversight. 

DITA SubjectScheme maps define consistent terms, define authoring consistency, and enable powerful navigation. It's not magic, but a structure, scalable solution. 

Without a structured taxonomy, content becomes the Wicked Witch, powerful but misunderstood. Customers struggle with findability, duplicate search results, and inconsistency. 

It takes time to properly classify content. Start small, then scale. Balance automation and human input. Governance is key. 

Formal taxonomy of subjects for products and features, user roles and personas, and other domains as needed. All managed by a corporate taxonomy group and applies to all content. It includes a formal change process and is accessible via REST API.

Markup requirements include capturing taxonomic classification in topics, using taxonomy ID to connect to master taxonomy, enabling reliable automatic updating of documentation source, and using built-in element types and attributes. An audit trail tracks classification actions over time.

A Journey for Your Content: DITA for Learning Environments

 Martina Schmidt and Barbara Kalous, of NINEFEB, stated that technical documentation is a high-quality, versatile information product that is more than a payment trigger and a must for policy compliance. It's support for service personnel and an information basis for training materials. 

Training materials are often inefficient. They last too long and little is retained. 

The vision is to break information silos of technical documentation and training. But the silos are very solid. So what does it take?

Fist, a shared data pool. Knowledge objects in a standardized XML scheme, such as DITA. The ability to manage media. Learning target formats are necessary. 

Theory without practice is sterile. Practice without theory is blind. 

DITA is good for creating structured technical documentation. Uses a topic approach for optimum reuse. Allows for a wide range of publications.  A topic contains all elements required to define a complete unit of information that addresses a single subject or answers a single question. Good for tech docs, but good for training? 

DITA 1.3 learning and training adds additional maps, interactions, and topic types specific for learning presentations.

Designing Curricular Transformation in Technical Communication Through Industry and Academia

 Rebekka Andersen, from the University of California, Davis, and Davis Carlos Evia, from Virginia Tech, started a multi-disciplinary research project to understand how content roles are involved in industry. 

It is a complicated love story between industry and academia in technical communication. But there's also a disconnect between industry and academia. They are not speaking the same language. And a relationship between two complicated entities will always be...complicated. 

Where are all the qualified college graduates? Al;so, perception that college graduates are not interested in the professions. TC in academia is "service' course, like a required course, undergraduate majors & minors, certificates, extension programs and advanced degrees. 

Many programs still do not offer instruction in structured authoring, content modeling, and other topics relevant in today's content world. 

Many students are unaware of the tech content discipline. Many also don't see "content" curriculum in their programs. Most students are not prepared to work in the content discipline. When students see content, they expect TikTok and Instagram, not DITA and structuring. 

Our kind of content goes by many names, including intelligent content, omnichannel content, technical content, and it is structurally rich, semantically aware, reusable, reconfigurable, adaptable, abnd rhetorically effective. 

Computer science is the future of academia, bit to the detriment of other disciplines. 

"Industry" is where he jobs are. What we call "content organizations." Managers report a disinterest in or lack of qualifications for working in highly technical environments. Entry-level content-focused position are becoming more hybrid, requiring background in both writing and programming. 

Delivering Targeted and Relevant Online Information

 Jonatan Lundin, of Excosoft, focused on content delivery and making sure content delivered is relevant. 

Relevant content delivery is being able to deliver the right content for the right role at the right time.

Products will require a "digital product passport" in the EU in the coming years to help deliver correct content for products in the future. 

Today's user has access to so much content, but most is not relevant. Delivering relevant content means hiding content that is not relevant to them. When users don't get content filtered, they have to find relevant content themselves, which means having to judge whether what they find is relevant and useful. 

Four user challenges when searching and reading:

1. Select a relevant information source.
2. Express the information eed to the selected source.
3. Assess the relevance of retrieved information.
4. Comprehend the retrieved information.

 So how to achieve relevant content delivery? An example is serial number-specific documentation. 

Content needs to be written and structured to support filtering. But it needs to be easy to understand. Structured authoring is the key. Content needs metadata. Content needs to be delivered in a portal that supports filtering.

The Power of Collaboration: Transforming Content Development for the Modern User

 Bravya Aggarwal, CEO & Founder of zipBoard, a content review and approval platform, started by saying that teams that collaborate effectively are 5 times more likely to achieve high performance. Often, content teams need to collaborate with people not on their direct team. Distance makes things even harder. 

Collaboration matters because user demand rapid, personalized, and engaging content. Collaboration can help meet that demand. Diverse--and expert--stakeholder expertise ensures insightful content. 

Challenges in content development include communication silos and disconnected workflows, slow and disorganized review cycles, and repetitive efforts. In addition, there is difficulty in managing feedback across multiple versions, confusion between internal and external feedback, and inefficient transition from feedback to actionable tasks. 

Good collaboration accelerates content production and refinement, encourages diverse perspectives, and aligns teams. 

Principles of effective collaboration include clear and centralized communication, asynchronous capabilities, and actionable feedback with clear accountability. Accountability happens in two phases, one by the end of the review, second when receiving feedback, as in who will make the fixes or changes. 

Important to make room for different working styles. 

AI has potential in automating routine tasks. It frees collaborators for strategic, creative input. But AI won't replace the role of people 

Tips to enhance collaboration include scheduling regular check-ins, implement clear feedback guidelines, and use visual aids and templates to standardize feedback.  Establish a dedicated review cycle with clear timelines and responsibilities, and document and share best practices and lessons learned, and do the latter across teams.

So Much Waste, So Little Strategy: Enterprise Content Strategy

 Scriptorium's Sarah O'Keefe kicked off one of ConVEx's 4 tracks noted that large enterprises don't typically have one source of truth for content across many buckets. 

Enterprise customer content is enabling content, information that enables people to do their jobs. The current state of enterprise technical content operations is technically challenging and often dangerous. Need a comprehensive solution for enterprise customer content. 

Challenges are culture, departments focus, and content silos. Culture is the biggest one. Tech comm and learning are often in different orgs. Tools for Tech Comm are optimized for technical content and learning management tools are optimized for learning content. A huge us vs. them problem. 

Why do we have content silos? They reflect corporate structure. Software is built to solution specific content types. Also, change is slow. 

There are legitimate differences. A knowledge base solve a specific technical problem. Many articles, but some are temporary. The audience is support. Learning consent is performance support.  Medium volume, audience is new users. Technical/product content is to get the people to use the product successfully, volume and longevity is high. (But just because a page doesn't get a lot of hits doesn't mean it's not important.)

With a CCMS, you assemble components into content objects (topics).  Use content objects to build deliverables. 

Even though chatbots are wrong and weird, customers like them because they are more fun that search. And also not filled with ads and other garbage. Chatbots are more satisfying, even though they are often wrong. (And everyone wants AI.)

You can't control your customers. You can't compete with ChatGPT. But you also can't control the AI. Your job is to make your information more appealing that getting an answer from ChatGPT.  

Search has a tendency to prioritize the critical mass. AI prioritizes differences. So your mistakes rise to the surface with AI.

Welcome Address and Opening Session

 Monday started early for several hundred people in San Jose as the Center for Information-Development Management's Dawn Stevens welcomed them to ConVEx 2025, the conference for content development and management professionals. She noted that the conference has grown since last year, with 70 folks from the local Silicon Valley area. 

The goal of the conference is to focus everyone on content strategy and content creation. The conference contains 76 concurrent sessions, as well as a "test kitchen." The agenda reflects the industry's on-going evolution, from writing well to reshaping content creation and delivery with AI. 

Creative change redefines content as user experience. Writers are now using design tools such as Figma and Canva. Progressive change is continuous upskilling in tech tools and using analytics to enhance content. Intermediating change is writers collaborating with designers and product managers and community-driven docs and participatory content. Radical change is AI-generated documentation and writers on product teams as UX/content designers.