Stream: Becketts
Time: 11:00 - 11:45
Every enterprise has a knowledge problem — and most don't yet realise it's the reason their AI projects are stalling. 90% of the operational expertise that keeps your mainframe running lives in the minds of a handful of senior engineers, buried in unindexed PDFs, scattered across email threads, or locked in decades-old runbooks that were never written for machines to read. When that knowledge gets fed into AI agents and RAG systems as-is, the results are disappointing: responses that are semantically plausible but operationally useless, agents that can quote documentation but can't reason like an experienced SME, and AI that your teams simply don't trust on high-stakes issues. This session introduces the concept of Knowledge Transformation — the critical, often overlooked step between "we have documents" and "we have AI that works." Drawing on real world scenarios and measurable impact, we'll explore: • Why unstructured enterprise content fails as AI input — and why traditional RAG alone can't fix it • The Knowledge Transformation pipeline: how raw content (videos, transcripts, PDFs, PPTs, tacit knowledge) is refined into structured, AI-consumable Knowledge Documents • how retrieval quality and knowledge density improves radically through Knowledge Transformation
There is currently no attachment for Your Mainframe Knowledge Is Not Ready for AI — Here's How to Fix It
Lewis is a software engineer embedded with the CICS Transaction Server for z/OS development team. Now within his 7th year at IBM, he joined as a Degree Apprentice, an early professional in the mainframe industry. Currently on secondment from the security team to work on IBMs mainframe AI mission.