Engineering Knowledge Management Systems

Building useful Knowledge Graphs will long be a Humans + AI endeavor. A recent paper lays out how best to implement automation, the specific human roles, and how these are combined. The paper, "From human experts to machines: An LLM supported approach to ontology and knowledge graph construction", provides clear lessons. These include: 🔍 Automate KG construction with targeted human oversight: Use LLMs to automate repetitive tasks like entity extraction and relationship mapping. Human experts should step in at two key points: early, to define scope and competency questions (CQs), and later, to review and fine-tune LLM outputs, focusing on complex areas where LLMs may misinterpret data. Combining automation with human-in-the-loop ensures accuracy while saving time. ❓ Guide ontology development with well-crafted Competency Questions (CQs): CQs define what the Knowledge Graph (KG) must answer, like "What preprocessing techniques were used?" Experts should create CQs to ensure domain relevance, and review LLM-generated CQs for completeness. Once validated, these CQs guide the ontology’s structure, reducing errors in later stages. 🧑⚖️ Use LLMs to evaluate outputs, with humans as quality gatekeepers: LLMs can assess KG accuracy by comparing answers to ground truth data, with humans reviewing outputs that score below a set threshold (e.g., 6/10). This setup allows LLMs to handle initial quality control while humans focus only on edge cases, improving efficiency and ensuring quality. 🌱 Leverage reusable ontologies and refine with human expertise: Start by using pre-built ontologies like PROV-O to structure the KG, then refine it with domain-specific details. Humans should guide this refinement process, ensuring that the KG remains accurate and relevant to the domain’s nuances, particularly in specialized terms and relationships. ⚙️ Optimize prompt engineering with iterative feedback: Prompts for LLMs should be carefully structured, starting simple and iterating based on feedback. Use in-context examples to reduce variability and improve consistency. Human experts should refine these prompts to ensure they lead to accurate entity and relationship extraction, combining automation with expert oversight for best results. These provide solid foundations to optimally applying human and machine capabilities to the very-important task of building robust and useful ontologies.

Every time you ask your AI a question, it forgets everything it learned the last time you asked. That's how most AI knowledge systems work today. They're called RAG systems — Retrieval Augmented Generation. You upload documents. The AI searches fragments on every query. Builds an answer from scratch. Closes the session. Next question? Same process. Zero accumulation. Andrej Karpathy — the person who built Tesla's AI vision system and co-founded OpenAI — published something quietly last week that might change this fundamentally (https://lnkd.in/dNB5swS8). He calls it "LLM Wiki." The concept: Instead of searching your documents every time, the LLM builds and maintains a persistent knowledge base. A structured wiki of markdown files with cross-references, concept pages, and contradiction flags. New document arrives? The LLM doesn't just index it. It reads it, extracts the key information, updates existing pages, notes where new data contradicts old claims, and strengthens the evolving synthesis. His key insight: "The tedious part of maintaining a knowledge base is not the reading or the thinking — it's the bookkeeping." Cross-references. Keeping summaries current. Noting when new data overrides old claims. That's exactly what nobody in your team has time for. And exactly what LLMs don't get bored doing. What this looks like in practice: Step 1 — Pick one knowledge domain. Not everything. One area where your team wastes time re-finding information. Customer onboarding. Product specs. Compliance requirements. Step 2 — Set up the structure. Claude Projects or Obsidian + Claude Code as the wiki layer. Raw sources go in one folder. The LLM-maintained wiki lives in another. Step 3 — Feed sources one at a time. Let the LLM summarize, cross-reference, and file. You review. Redirect. Ask follow-up questions. The wiki grows with every session. Step 4 — Query against the wiki, not the raw documents. Answers are faster, more contextual, and cite specific pages. The open-source project Graphify already implements this pattern — claiming 70x fewer tokens needed to answer questions compared to raw-folder RAG. For now, this is still a developer concept. No plug-and-play enterprise solution exists yet. The benchmarks are anecdotal, not peer-reviewed. The wiki could drift or hallucinate if not curated. But the direction is clear: AI that builds knowledge, not AI that searches it. Save this for when your team's knowledge base fails you again. — 📌 Save this post for later ♻️ Share it to inspire your network Follow Hartmut Hübner, PhD for AI insights that work.

Engineering Data is everywhere: distributed across tools, documents, and platforms. But if we want AI to truly understand our products and support development, we need to link that data, create traceability, and make it accessible and modifiable. At Accenture together with Vlad Larichev and many other colleagues, we see 3 powerful approaches to unlock engineering data with AI: 1️⃣ Retrieval Augmented Generation (RAG) Aggregate your distributed engineering knowledge into a Vector Database. This enables semantic search and document-level Q&A across tools and documents. When paired with a Language Model (LM), relevant context is retrieved based on similarity to your prompt, perfect for engineering Q&A and documentation support. 2️⃣ Graph Retrieval Augmented Generation (GraphRAG) Link your engineering data across domains using a Knowledge Graph. Capture relationships between requirements, CAD, simulation, test data, etc. Enabling traceability and holistic V-model understanding for your LM. Essential for typical cross-domain tasks, such as impact analysis and E2E configuration management. 3️⃣ Model Context Protocol (MCP) Why move your data at all and store it in additional databases? With MCP, AI agents can access and modify data directly inside your tools without data ingestion and storage efforts. It’s an agentic interface to your engineering stack, which enables cross-domain data access, retrieval and generation, making it suitable for E2E ECR processing. These aren’t just technical solutions. They’re a paradigm shift in how we will interact with engineering data and develop complex products. These methods only unlock their full potential when high-value use cases are identified and applied in a goal-oriented way. Interested in how this can work for your organization? Let’s talk. — Dr. Matthias Ziegler | Dr.-Ing. Tobias Guggenberger | Arne Breitsprecher | Georg Brutzer | Florian Böhme #EngineeringIntelligence #DigitalEngineering #ProductDevelopment #Accenture

𝗜𝗳 𝗸𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 𝘄𝗮𝗹𝗸𝘀 𝗼𝘂𝘁 𝗼𝗳 𝘆𝗼𝘂𝗿 𝗽𝗹𝗮𝗻𝘁, 𝘀𝗼 𝗱𝗼𝗲𝘀 𝘆𝗼𝘂𝗿 𝗰𝗼𝗺𝗽𝗲𝘁𝗶𝘁𝗶𝘃𝗲 𝗲𝗱𝗴𝗲. For years, the most valuable lessons in manufacturing lived in people’s heads, sat in spreadsheets, or got buried in reports no one read. KM systems existed — but insights stayed siloed. Maintenance logs didn’t inform production. Quality data never reached design. And when experts left, their know-how left too. KBE helped. Rules were codified, designs automated, tasks accelerated. But it stayed narrow. Useful for engineering but useless for the wider enterprise. Now Smart Factories are rewriting the rules. The “smart” in Smart Manufacturing is no longer just IoT, AI, or digital twins — it’s Knowledge Management evolving into the core ingredient that makes factories adaptive, resilient, and truly smart. This transformation is a true 𝗛𝘂𝗺𝗮𝗻–𝗢𝗿𝗴𝗮𝗻𝗶𝘇𝗮𝘁𝗶𝗼𝗻–𝗧𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝘆 (𝗛𝗢𝗧)  shift: • Organizations evolve into networks that learn. • Employees grow into knowledge partners. • Technology connects, structures, and scales those learnings. 𝗔𝘁 𝘁𝗵𝗲 𝗼𝗿𝗴𝗮𝗻𝗶𝘇𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗹𝗲𝘃𝗲𝗹: From hierarchies to networks. Lessons flow horizontally across production, logistics, and supply chains — and vertically into ERP and PLM, the true nerve centers. PLM connects design and engineering with shop-floor feedback, ensuring learnings inform product evolution, not just production routines. The enterprise itself becomes a learning system as every captured lesson strengthens resilience, speed, and customer value. 𝗔𝘁 𝘁𝗵𝗲 𝗲𝗺𝗽𝗹𝗼𝘆𝗲𝗲 𝗹𝗲𝘃𝗲𝗹: From operators to knowledge workers. Every workaround, every fix, every idea feeds the system. From machine cooperation to human–machine collaboration. Cobots and AI extend capability, people bring judgment. From one-time training to continuous learning. KM guides workers in real time, embedding best practices into daily decisions. 𝗔𝘁 𝘁𝗵𝗲 𝘁𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝘆 𝗹𝗲𝘃𝗲𝗹: 𝗖𝗮𝗽𝘁𝘂𝗿𝗲 – Every event, fix, and insight is logged with context. 𝗘𝗻𝗿𝗶𝗰𝗵 – AI structures it, links it, and connects it with past cases. 𝗜𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗲 – Learnings move across functions and levels. 𝗔𝗽𝗽𝗹𝘆 – Best practices feed back into workflows, tools, and training. 𝗘𝘃𝗼𝗹𝘃𝗲 – Each cycle makes the system smarter. Documentation, once a burden, is now the nervous system. It captures memory, creates best practices, and feeds them forward — so tomorrow’s decisions are always better than yesterday’s. In Smart Manufacturing, knowledge isn’t just an asset but the very definition of smart. Winners will be those who capture lessons, turn them into best practices, and scale them across the ecosystem. 𝗘𝘃𝗼𝗹𝘂𝘁𝗶𝗼𝗻 𝗶𝗻 𝗼𝗻𝗲 𝗹𝗶𝗻𝗲: Tribal Know-How → Scattered Learnings → Codified Rules → Knowledge-Enhanced Practices → Ecosystem-Wide Best Practices Ref:https://lnkd.in/dTiqaCQu

Much of the conversation about enterprise AI focuses on model quality, prompt engineering, and optimization. These are real problems. The engineering challenge I spend most of my time on is less discussed and more consequential. Entity resolution across heterogeneous enterprise data sources. The same person appears in at least six different forms across a typical executive's work stack: a name in email, an email address in calendar, a nickname in Slack, a full name and title in CRM, a display name in Zoom, and some variation of all of the above in shared documents. The same project might be "Q3 initiative" in one meeting, "the Atlas project" in another, and "that thing we decided in August" in a third. If you can't resolve these into a unified, consistent representation, you are not building a knowledge system. What you are building is a fast way to retrieve fragments that may or may not refer to the same things. Fun facts to know and tell. A knowledge-centric architecture depends on getting these things straight. A model of an executive's work is only useful if the entities that model references are correctly identified and consistently maintained. A commitment attributed to the wrong person, or a project split across three different representations, means the system is reasoning over a flawed picture. The LLM will produce confidently incorrect output based on a flawed picture. The executive will quickly lose trust. Twenty-five years of building enterprise systems taught me this: the unglamorous work of data modeling, entity resolution, and pipeline reliability are where trust is won or lost. The demo shines with its pristine data. Production data is a different planet. Entity resolution is the prerequisite for everything else in a knowledge-centric architecture. It's where the deepest defensibility comes from. When a system is able to correctly resolve your professional world into a structured graph, that system won't easily be replicated by competitors starting from scratch.

One of the evergreen promises—and persistent challenges—of knowledge management in AEC is scaling expert knowledge. In a business built on expertise and results, firms need to help emerging professionals become more effective healthcare architects, bridge engineers, project managers, or sustainability experts. Historically, that kind of learning has happened slowly: one project at a time, one mentor at a time. But today, the urgency is rising. Many firms are facing a real-time talent crunch: senior experts are retiring, hiring is competitive, and there’s pressure to ramp up new staff faster than ever. That’s where KM teams are stepping in—not just to collect or store knowledge, but to accelerate the transfer of critical insight from experts to emerging professionals. One increasingly popular method is structured expert interviews. Here's an example: https://lnkd.in/guFC-9CR Instead of asking a senior architect to write down everything they know (which rarely works), KM teams are capturing informal conversations on video, then using AI tools to transcribe, summarize, and transform them into searchable, reusable knowledge. In some cases, those interviews also form the basis for courses or training programs. These interviews take different forms. Sometimes they’re one-on-one conversations recorded in a conference room or over Zoom. Other times, they’re held as live learning events—inviting staff to listen in, ask questions, and absorb the exchange in real time. In some firms, experts are even interviewing one another, creating space for reflection and storytelling while modeling a culture of shared learning. Regardless of format, the resulting videos become assets that can be reused across onboarding, training, and AI-powered search. KM teams act as knowledge brokers, working across departments and generations to extract tacit expertise and make it available in the flow of work. That includes surfacing bite-sized insights through search, packaging repeatable methods into standards, or embedding lessons learned into onboarding programs. AI is playing a major role throughout this process—making transcription and summarization faster, enabling retrieval through tools like AI search, and helping turn raw insights into new knowledge assets for experts to review. But the shift is as much cultural as it is technical. The best KM teams are creating lightweight, repeatable ways to scale expert knowledge without putting all the burden on the experts themselves. 💡 This is Trend 8 of 12 from the Issue 6 of Smarter by Design: “How Leading AEC Knowledge Management Teams Are Evolving to Thrive in the AI Era.” 📖 👉 Read the full issue: https://lnkd.in/gYgrFzVN #AEC #KnowledgeManagement #SmarterByDesign

Outdated knowledge assets aren't just an administrative issue—it's a safety and quality issue that affects everyone. Growing concerns and the loss of experienced personnel are creating more and more pain oints across manufacturing. Many recognize that knowledge management must be a strategic focus for accelerating workforce competency, protecting operational performance, and driving innovation. But moving from a strategic focus to embedding it within the cultute of an organization is where it gets tricky. In my conversations with manufacturing leaders, organizations that are most successful at maintaining current documentation have gone deeper than just building systems—they've created cultures that genuinely value knowledge sharing and accuracy. One way to support this cultural shift is by creating a "knowledge health index"—a dashboard showing the status of knowledge assets across operations. Some of The most effective systems include: 🔁 Collecting feedback from users on documentation quality and accuracy, in a non-intrusice way as soon as the job is done. 📈 Monitoring usage patterns to identify which procedures are most frequently accessed 🫶Mature capabilities that can automatically flag documentation that may need updates based on system changes 🤝Create a way to gauge how often knowledge assets are being reviewed and updated - and recognizing those who lead by example. One manufacturer who implementing this approach has maintained over 95% accuracy in their critical knowledge assets, compared to less than 60% before implementation. The difference between organizations that struggle with knowledge management and those that leverage knowledge to create value isn't just adopting technology—it's creating an environment where everyone understands that knowledge is a shared asset that requires collective stewardship. The LNS Research Industrial Knowledge Management framework provides a scalable approach for manufacturers, regardless of where you are in your journey. What approaches have you seen work for maintaining critical operational knowledge in your organization? 📣 I'd love to hear your experiences in the comments. ⬇️ #KnowledgeManagement #IndustrialTransformation #OperationalExcellence #ConnectedWorkforce #DigitalTransformation

Your engineers are brilliant. That's why they keep solving the same problem at different facilities. Over and over. Without knowing someone already figured it out. This isn’t an intelligence problem. It’s an infrastructure problem. Plant A has brilliant engineers.  They found a quality issue costing $8K a month.  Spent three weeks finding the root cause.  Built a smart solution.  Saved $100K a year.  Documented everything.  Problem solved. Six months later, Plant B found the same issue.  Did the same analysis.  Built the same solution.  Saved the same $100K.  Documented it separately.  Problem solved again. Plant C? Also brilliant.  They’re discovering the same issue right now.  Starting the same process.  They’ll solve it soon, for the third time. Same company.  Same brilliance.  Zero knowledge sharing. Each plant keeps its own notes.  No central system.  No easy search like “Has anyone solved this before?”  No alerts when similar problems show up.  No way to turn local wins into company standards. So every plant starts from scratch.  And your best practices stay trapped. Spreadsheets on local drives.  Old email threads.  PowerPoints buried in folders.  Knowledge stuck in people’s heads. Hundreds of great ideas are locked away.  While others waste time reinventing them. That’s lost time, lost money, and lost progress. The best manufacturers treat knowledge like inventory.  You wouldn’t let one plant hoard materials while another runs short.  So why let one plant hoard solutions? When Plant A solves something, it should go into a shared system.  Tagged by equipment, process, and problem.  Searchable for everyone.  Alerting others when similar issues appear.  Scalable across all plants. That’s how local wins become company standards. Plant A’s $100K idea  becomes $300K when shared with B and C.  Same effort.  Triple the impact. In three weeks, all plants could be aligned, instead of six months of duplicate work. Your engineers stop reinventing and start innovating.  New engineers learn faster.  The whole company gets smarter. You already have brilliant engineers.  You already have brilliant solutions.  Now it’s time to multiply that brilliance, not trap it. Because every month knowledge stays isolated,  your competitors move ahead.  They’re solving once and scaling everywhere. Your engineers are brilliant.  Your solutions are excellent.  Your knowledge sharing is broken. Fix the infrastructure,  and brilliance multiplies. P.S. If your best practices are trapped on islands,  let’s talk about building the system that sets them free.  DM me “KNOWLEDGE.”