VIDEO PODCAST
OpenText's Savinay Berry: Context Engineering Drives Enterprise AI
Terry Riley
January 1, 2026

In this episode of Tech Voices, I spoke with OpenText Executive Vice President and Chief Product and Technology Officer Savinay Berry joins host James Maguire to examine why context engineering has become the critical differentiator in enterprise AI. The discussion cuts through agentic AI hype to focus on the practical realities enterprises face: where meaningful data actually lives, how trust and security shape adoption, and why AI initiatives fail without deep contextual grounding. Drawing on OpenText's decades-long role inside the enterprise firewall, Berry explains how organizations can move from experimentation to real, repeatable ROI by aligning data, agents, and governance.

Core Takeaways
Enterprise AI rises or falls on context. Without access to high-qualit
Enterprise AI rises or falls on context. Without access to high-quality, well-governed enterprise data, AI agents lack the situational awareness needed to deliver reliable, business-critical outcomes.
The most valuable data in the world is not public. While the open inte
The most valuable data in the world is not public. While the open internet contains tens of zettabytes of information, enterprise systems hold hundreds of zettabytes of proprietary data that remain largely untouched by AI models.
Context engineering requires transforming data, not just collecting it
Context engineering requires transforming data, not just collecting it. Raw documents and files must be converted into data products and knowledge graphs that reflect relationships, history, and operational decision-making.
Real ROI comes from starting small and scaling deliberately. Enterpris
Real ROI comes from starting small and scaling deliberately. Enterprises see measurable returns when agents are deployed against focused use cases—such as IT operations, supply chain tracking, or claims processing—and expanded only after trust and accuracy are proven.
KEY QUOTES

The Data Advantage Behind the Enterprise Firewall

The publicly available data on the internet, even if you include everything from Wikipedia to major news publications and public libraries, is about 10 to 15 zettabytes. That's a lot of data, and it's what enabled the models we have today. Without the internet, we wouldn't be here." "What's far more interesting is the data sitting behind enterprise firewalls. Across hundreds of thousands of companies, that data is closer to 150 to 200 zettabytes. It hasn't been indexed at scale, and accessing it requires years of trust and security relationships. That's where the next wave of enterprise AI value will come from.

Why Context Determines Whether AI Works or Fails

If you don't have the data, you don't have the context, and enterprise AI is going to fail. Context engineering is about making sure an agent understands not just a document or a rule, but the full history and pattern of decisions that led to outcomes." "You can't just throw spreadsheets or PDFs at an agent and expect value. Data has to be pre-processed into data products and knowledge graphs that capture how documents, objects, and decisions connect. That's what enables agents to act intelligently and autonomously.

From AI Anxiety to Secure, Governed Agents

A major concern executives raise is data leakage. They want to build agents, but they worry that once proprietary data goes into a model, they lose control over where it ends up. That concern is completely valid." "That's why enterprises need an end-to-end AI data platform. It has to manage the full lifecycle—from data access and preparation to secure runtime and identity controls—so organizations know exactly how data flows and how agents are governed.
ABOUT THE AUTHOR
Terry Riley
Editor