AI agents are the newest consumers of analytics, and they don’t have the instincts a human analyst does. When a person sees a number that looks wrong, they know to ask questions. An agent doesn’t pause. It takes the answer it gets and asks the next question, and the next, compounding any mistake along the way.
In this on-demand webinar, Donald Farmer, TDWI Research Fellow, shares new TDWI research on AI governance, data trust, and the state of the semantic layer. He sits with Dave Mariani, founder and CTO of AtScale, to talk through what’s actually stopping most organizations from getting there.
Even the best LLM on the planet, handed the full schema and every column definition, gets real business questions right only 24% of the time on the BIRD benchmark, a conversational test built to mimic a real conversation between a business user and an AI agent. Follow-up questions compound that error, no matter how smart the model is.
Watch the on-demand session to see why two-thirds of organizations call a semantic layer critical, why fewer than one in five have one, and what’s standing in the way.
What You’ll Learn
- The gap between planning multi-agent AI and running it: 38% are exploring or planning multi-agent systems, only 26% have deployed one
- Why AI governance lags data governance today (47% call their AI governance immature versus 36% for data), and which one you need to fix first
- Fewer than half of organizations trust their own structured data, and only about a quarter have a real semantic layer in place
- What separates a semantic layer from a glossary: an engine that computes the same answer every time, not just a description of what a metric means
- How leading teams treat semantic models as code: version-controlled, reviewed through pull requests, and tested continuously against a set of certified answers
Why Watch
Every organization in TDWI’s research wants AI agents that can be trusted to act on real business data. Almost none of them agree on how to get there, and TDWI’s research points to why: data silos and data quality are the problem most teams brace for. Solve those, and a second one is waiting right behind it: inconsistent metadata and a semantic layer that was never built.
Donald Farmer walks through what TDWI’s research says is actually holding teams back, and it usually isn’t cost or technology. Then Dave Mariani, who’s spent 13 years building a universal semantic layer at AtScale, gets specific about why the real unit of trust is a semantic engine, not a glossary or a catalog.
Speakers
Donald Farmer Principal, Treehive Strategy, and Research Fellow, TDWI. A data and analytics strategist with more than 35 years in the field, Donald has led product design and innovation at both Microsoft and Qlik, and now advises technology vendors, startups, investors, and government organizations on data, analytics, and AI strategy.
Dave Mariani Founder and Chief Technology Officer, AtScale. Prior to AtScale, Dave was VP of Engineering at Klout and at Yahoo, where he managed some of the world’s biggest data and analytics challenges. He’s a big data visionary and serial entrepreneur who’s spent more than a decade building AtScale’s universal semantic layer.
