Steve Wooledge

Running an AI pilot is easy. Production is where it gets real: costs you cannot predict, and answers the business will not trust. Scout24 has run AI agents on production data for a year. Angelita Frozza Sanches, Head of Data, walks through what her team built and learned. A context layer that gets richer as people and agents use it. Kept portable, so knowledge is not tied to one model or platform. Guarded, so an agent sees only what the person behind it can see. Answers drawn from their curated query library carry a 100% confidence score. Mike Ferguson advises global enterprises on data and metadata strategy. He asks whether a single context layer is reachable once an estate has fragmented, and what it takes to keep meaning in sync when systems model metadata differently. Suresh Srinivas has built metadata systems three times, at Hortonworks, Uber and Collate. He covers what makes the approach repeatable, and what benchmarks show for AI accuracy and token cost. You will leave knowing how to improve AI accuracy on your own data, stop agents running expensive queries, and what to build first if you are starting from nothing. For data teams, architects and AI engineers.