In this episode, Alex Hayes is joined by Stephen Nwoye following his presentation at a Data Community meetup. They explore what it takes to move beyond basic retrieval-augmented generation (RAG) and build AI systems that truly understand organisational data.

Stephen explains how curated vector databases, summarised context layers, and strong governance frameworks help control what AI can access, improve relevance, and protect sensitive information. The conversation also covers the practical challenges of implementing governance in large organisations, why starting with a primary dataset is key, and how building a central knowledge repository improves trust, speeds up development, and reduces reliance on tribal knowledge.

They also discuss testing approaches, documentation at scale, and why owning your organisation’s knowledge layer unlocks future AI opportunities — even without direct access to live data.

Topics covered:

Why governance matters more than just connecting AI to documents
Using vector databases as a curated knowledge layer
Starting small with a primary dataset
Building trust through documentation and transparency
Reducing tribal knowledge and speeding up onboarding
Testing AI-assisted data development
Owning your organisation’s AI-ready knowledge repository