Don Murray

LLMs are becoming a commodity. Organizations will run several, and the cost of running them is already a real concern. What separates a toy from a useful system is context, and great context requires all the data: the easy data and the hard data, the geospatial, the legacy, the information locked behind a firewall. Model Context Protocol is becoming the standard way to give agents that context, but most implementations stop at clean, cloud-hosted APIs. Real enterprises don't run on clean data alone. After 30 years connecting the world's messiest data, Don Murray covers what it takes to bring MCP to the data enterprises have, not the data vendors wish they had. Where geospatial and legacy data break AI tooling assumptions. How to deploy across cloud, on-premises and hybrid. And what security teams need before they sign off on an AI touching production systems. Then three live demos on one FME MCP server, reaching the 1,000-plus systems FME reads. First, Snowflake’s LLM reaches data that never made it into the warehouse. Next, Databricks doing the same on a different stack. Finally, showing the same thing with Claude. Everything shown is available today and nothing on stage you could not run today. FME Flow's MCP server is available now. Same MCP server, three clients, no rework. That keeps the model layer a commodity you can shop on price, and stops you copying the same data into every platform that wants to reason over it.

Most AI initiatives can reason, but few can act. Your Databricks lakehouse has the data, but not the enterprise connections to put AI to work. Without them, AI remains a passive chatbot, not an autonomous agent. Bridging this gap has meant writing brittle, hard-to-scale glue code. MCP and FME solve this with a blueprint for a scalable AI Tool Factory. Safe Software CEO Don Murray shares how FME uses MCP to transform complex workflows into plug-and-play tools for any AI. You’ll learn how to: - Eliminate technical debt: Deploy pipelines into governed MCP tools without writing code. - Bridge the data gap: Connect AI agents to spatial data, legacy systems and real-time feeds the lakehouse can’t easily reach. - Deploy at scale: Quickly build, test and ship AI tools. - Actionize the lakehouse: Turn static insights into autonomous enterprise actions. Learn the all data, Any-AI, MCP-enabled strategy to turn your Databricks platform into an actionable AI powerhouse.

AI agents are only as powerful as the data they can reach. Most MCP tools connect to the easy data. But enterprise value lives in spatial, CAD, BIM, legacy databases, and real-time feeds — systems that will never have a native MCP interface. FME gives organizations an All Data MCP Server: any workflow becomes an MCP tool, on any data, available to any AI. 500+ systems. Thousands of possible tools. Cloud and on-premises. No exceptions.

Enterprises are embracing generative and agentic AI as models evolve faster than ever, creating uncertainty. The Model Context Protocol (MCP) resolves this by standardizing connections between AI and underlying services. We demonstrate how the ability to both consume and build MCP services provides...

Disconnected data is a death sentence for the enterprise. While you struggle to connect all your data to AI, competitors that have connected all their data to AI are making decisions that will put you out of business. To survive, you must connect all data to any AI. The Model Context Protocol (MCP) provides this foundation. Through live demos, we show how to build and consume MCP services that turn all your data into a competitive advantage.

In today’s world, the organizations that succeed are the ones that can pivot quickly, confidently, and without disruption. The power of choice is everything. Today organizations must be able to choosing which data to use, and systems to use, and how to adapt when priorities shift is now critical for survival. The West Yorkshire Combined Authority (WYCA) faced a high-stakes challenge: prepare for multi-million-pound bus franchise negotiations with clear, defensible evidence of network performance. Using FME, they transformed over 2.5 million real-time GPS bus pings into a scalable, grid-based value-scoring model that revealed exactly which routes and operators delivered the most value. The result? Faster, in-house analysis without costly consultancy fees, stronger negotiating power backed by objective data, and a repeatable blueprint for data-driven ROI that is applicable across the UK public sector and beyond. Join Safe Software and WYCA to discover how all-data, any-AI integration can turn raw, real-time data into actionable intelligence empowering organizations to make smarter decisions, faster.