Letting LLMs Operate, Safely: What Matters When Building an AI Architecture for Water Utilities

November 12, 2026
Time: 11:00 AM ET | 10:00 AM CT | 8:00 AM PT
Duration: 60 minutes
Already registered? Click here to log in now.

Summary

Connecting an LLM to a water utility’s data is the easy part. The real challenge begins when AI moves beyond simply reading and interpreting information and starts acting on operations.

Making an LLM interact safely with operational systems, run agentic workflows autonomously, and support a utility’s operations end-to-end — rather than addressing a single vertical use case — requires a fundamentally different approach to AI architecture.

At Xylem Vue, we have spent the last few years building for exactly this challenge. In this webinar, we will share the lessons learned along the way and explore what it takes to move from AI experimentation to solutions that can operate safely within real water utility environments.

Through live examples based on real water operations, we will look at the journey from reading sensor data to enabling autonomous agentic workflows. We will also explore the different levels of AI maturity in water utilities, why architecture must leave room for each utility’s own operational know-how, and what utilities should demand from AI solutions designed to become part of their day-to-day operations.

What you’ll learn:

• Why reading and acting are two different AI problems.
• Live examples, from reading sensors to autonomous agentic workflows.
• The different levels of AI maturity in water utilities.
• Why AI architecture must leave room for the utility’s own know-how.
• What water utilities should demand from any AI solution.