Resource Library : DATA SCIENCE CONNECT WEBINAR
The Single Source of Truth: Building Data Foundations for Agentic AI
Mark Palmer, Chief Strategy and Marketing Officer at AtScale, joins leaders from SAP, Skyflow, and Turgon on what an AI agent needs before you can trust its answer.
For decades, the single source of truth meant a database, with a human analyst catching the bad join and knowing which revenue definition the CFO meant. Agents don’t have that analyst. When they hit a gap, they guess.
In this on-demand panel from Data Science Connect, Mark argues that for AI, the source of truth lives in the meaning: the business definitions and calculations an LLM can’t work out from a schema. On the BIRD-Interact benchmark, he notes, the best LLMs handed raw enterprise data top out around 38% accuracy.
What You’ll Learn
- Why the source of truth is moving from the database to a single source of meaning
- Why an LLM still guesses even with clean data, and what the BIRD-Interact results show
- How to test agent accuracy with evals built from your own questions and KPIs
- How to give agents data safely, with runtime controls and permissions that travel with the data
Why Watch
Ask an agent for revenue and it has to pick a definition. One AtScale customer, Blue Yonder, has 20 of them. Mark explains how AtScale defines each metric once, computes it the same way for every agent, and tests its accuracy against your own questions. SAP, Skyflow, and Turgon add business process context, data protection, and how to keep context current.
Speakers
Preeti Vaidya, Executive Director, Data and AI, Morgan Health (JPMorgan Chase). Moderator.
Mark Palmer, Chief Strategy and Marketing Officer, AtScale
Ankur Varma, CEO and Co-Founder, Turgon
Amruta Moktali, Chief Product Officer, Skyflow
Shubham Trivedi, Digital Solution Advisor, SAP
Hosted by Jenna Holland, VP of Marketing and Digital, Data Science Connect.