OCLC research on AI in cataloging: addressing backlogs and workflows

I am very pleased to share that the first in a series of four blog posts is out. The post titled "Backlogs and beyond: AI in primary cataloging workflows" discusses how libraries are addressing cataloging backlogs, legacy metadata issues, and language challenges by exploring the practical support AI can provide, while keeping human expertise central. This work originated from discussions in the OCLC Research Library Partnership Metadata Managers Focus Group and the subsequent working group that has emerged from these ongoing conversations over the years. Many thanks to those who contributed to this effort. I look forward to sharing more in the series and engaging in a conversation about the work. Read more here: https://lnkd.in/eC_6a73i

The third blog post in our AI and metadata workflows series is out -- this one is written by my colleague Brian Lavoie and focusses on workflows for institutional repositories! Read and find out -- can AI help to manage institutional repository metadata? From improving deposit workflows to cleaning up legacy data, explore how AI might support IRs while keeping human expertise front and center. Our final blog post will be out next week! Stay tuned. https://lnkd.in/gMaQjTyy

AI and the Untapped Potential of Content My further thoughts from Hyland's CommunityLive event (part 2) 𝐀𝐈 𝐚𝐬 𝐭𝐡𝐞 𝐅𝐮𝐭𝐮𝐫𝐞: Even though the generative AI bubble may be deflating, AI itself is here to stay. Hyland’s new platform aims to exploit massive stores of unstructured data (content), which is crucial for enterprise AI. 𝐔𝐧𝐭𝐚𝐩𝐩𝐞𝐝 𝐎𝐢𝐥 𝐑𝐞𝐬𝐞𝐫𝐯𝐞𝐬: Hyland and similar companies are sitting on huge content "reserves," but they face the challenge of convincing customers to invest in extracting and using this data. The opportunity is massive, but it requires time and resources to capitalize on. 𝐓𝐡𝐞 𝐑𝐨𝐚𝐝 𝐀𝐡𝐞𝐚𝐝 𝐟𝐨𝐫 𝐇𝐲𝐥𝐚𝐧𝐝: While Ghai’s strategy makes sense and has been well-received, the real test will be delivering on these promises. Hyland has a history of making big announcements that didn’t materialize, so this time they must prove they can execute. 𝐄𝐧𝐝-𝐔𝐬𝐞𝐫 𝐁𝐮𝐲-𝐈𝐧: End-users remain a key hurdle. If they see practical benefits, like using AI without moving between platforms, they may finally engage with Hyland’s new vision. 𝐓𝐚𝐤𝐞𝐚𝐰𝐚𝐲: Hyland’s AI-driven future holds immense potential, but its success depends on its ability to convince customers that investing in content extraction is worth the effort. Don't forget to follow us here or head over to our website to subscribe to our blog for regular industry thoughts and insights. #AI #genAI #ECM #Hyland

EN Just finished reading the latest book from Ole Olesen-Bagneux, Fundamentals of Metadata Management, and it's mind-blowing. In an era of AI, metadata is crucial for success, and managing it is very hard. The main point in the book, at least for me, is that it is not about a tool. It is not about a framework(or is it?). It is not about data catalogs. It is about a metadata CULTURE. Everyone, everywhere in an enterprise, is responsible for some part of metadata. Organizing it is essential and must be done right away. It can be as small as a spreadsheet and may already exist! It just needs to be recognized as such an important thing. One takeaway from all I've been reading about metadata is that it must be a whole-enterprise adventure. Ole's book brings the discussion to a broader audience, as I think that it is easy for non-data and even non-technical people to understand and grasp. #metadata #management #data #AI

Big (or small) news here: Anthropic released Claude Haiku 4.5 today, and this small model matches the performance of their mid-sized Claude Sonnet 4 (released in May) at one-third the cost and more than 2x the speed. For small and medium-sized businesses, AI budgets have been a major bottleneck. With Haiku 4.5, Anthropic says, it's now actually economically viable to offer AI agents and chatbots to free-tier customers. Anthropic says that the best way to use Haiku 4.5 is to first have Sonnet 4.5, their current flagship model, handle strategic thinking and complex planning. Then, you spin up multiple instances of Haiku 4.5 to execute the actual work in parallel. In coding, that means Sonnet plans the build while multiple Haikus tackle different components simultaneously. In finance, Haiku monitors thousands of data streams while Sonnet does the deeper predictive modeling. Gamma tested Haiku 4.5 on slide generation and hit 65 percent accuracy. Their previous best model? 44 percent. Their founder called it a game-changer for unit economics. The new model launch is a clear sign of how Anthropic plans to stay competitive in the AI race. Frontier AI stays pricey, but cheaper alternatives get smarter every release. For founders bootstrapping their way forward, that's everything. https://lnkd.in/eGmhdUiF #AI #StartupEconomics #Claude #ArtificialIntelligence

In a world driven by AI, your data is more than just ones and zeros—it’s the foundation for innovation. Two of my favorite people, Jason Lohrey, founder of Arcitecta, and Dell’s Alex Timbs, recently sat down to discuss the evolving role of data in the age of AI. Jason shared some incredible insights about why metadata, context, and human intent are more critical than ever. From treating data as a true asset to enabling real-time, low-latency access with solutions like Arcitecta’s Mediaflux Real-Time®, their conversation highlighted how technology can empower people and organizations to innovate and thrive. This is about more than just technology—it’s about unlocking new possibilities. Read the full article to learn more. https://del.ly/6045AxR27 #AI #DataManagement #Arcitecta #iwork4dell

🤖 GenAI is moving fast, but without semantics, it won’t scale. Enterprises are finding that LLMs alone can’t guarantee accuracy, trust, or compliance. Without a semantic layer, “revenue” might mean three different things to three different AI systems. In his latest blog, AtScale SVP Cort Johnson breaks down why semantic layers are now the foundation for enterprise GenAI success, and why GigaOm named AtScale a Leader + Fast Mover in its 2025 Radar Report. 📖 Read the blog: From Models to Meaning: Why Semantic Layers Are the Foundation of Enterprise GenAI Success 👉

We often talk about making our content AI-ready, but what does that mean at the file format level? In my latest post, I break down how Markdown, JSON, YAML, and XML perform when humans read and AI retrieves: 📘 Markdown – clear, structured, and easy for both 🧩 JSON – precise and machine-friendly 🧱 YAML – readable with just enough structure ⚙️ XML – reliable, but built for another era My takeaway: Markdown combined with YAML or JSON metadata strikes the best balance between human readability and AI structure. Would love to hear what format does your team use most often? #AIReadyContent #TechnicalWriting #MicrosoftLearn #GoogleDocs #KnowledgeAgent #ContentStrategy #Markdown

Everyone’s talking about LLMs — but without context, they hallucinate. That’s why **RAG (Retrieval-Augmented Generation)** with **Vector Databases** matters. Instead of forcing a model to “remember everything,” RAG lets it *look things up* in real time. 🔹 Accurate → grounded in your own data 🔹 Efficient → no need for massive fine-tuning 🔹 Scalable → updates instantly when knowledge changes 💡 My takeaway: RAG + Vector DBs are becoming the new baseline for enterprise AI. If you’re not learning this stack, you’ll be behind. 👉 Are you already experimenting with RAG pipelines in your projects? #AI #RAG #VectorDatabase #MachineLearning #LLM #DataScience

New Post: Have LLMs overtaken humans for knowledge organization? - https://lnkd.in/e-w6EiEa - If they have, you might not have thought so from the ISKO-UK annual conference (7 October 2025). The theme for this online conference was “knowledge organization in the age of AI”. For those unfamiliar with the terrain, “KOS” refers to “knowledge organization system”, a catch-all term that comprises controlled vocabularies, taxonomies, thesauri, and the like. […]