Speakers: Stephen Ramos (PG&E), Siny George (PG&E), Cindy Chang (PG&E)

Abstract: 

Generative AI is changing how operational software is developed, but many organizations are still determining where the technology delivers measurable value. This session presents PG&E's experience using Claude through Kiro as part of its digital transformation journey, exploring how AI-assisted development can accelerate the delivery of operational applications that support gas control, pipeline operations, and regulatory compliance.

Before writing a single line of code, PG&E's Gas AI Task Force applies Lean principles to decide where AI actually fits. Through structured problem-solving sessions and value stream mapping with process owners, the team maps how work really happens, surfaces bottlenecks and inefficiencies, and pinpoints the specific steps where AI can add value versus where it can't. Every idea then runs through a consistent scoring model — weighing value (does it move the business forward?), feasibility (do we have the data, people, and buy-in?), and fit (will it plug into how work already happens?) — so ideas are ranked on clear, repeatable criteria rather than gut feel. This front-end discipline builds a defensible pipeline of high-impact use cases and keeps the team focused on solving real problems, not chasing technology for its own sake.

Through several real-world projects, developed rapidly by our technology team combining deep Gas operational data knowledge with AI-assisted software development, the solution illustrates how organizations can use AI to unlock more value from their existing SCADA and enterprise systems—moving from manual data searching toward intelligent exception management. The presentation discusses development workflows, productivity gains, challenges encountered, validation strategies, and key lessons learned along the way.

Rather than presenting a theoretical view of AI, this session offers practical examples and actionable guidance for utilities and pipeline operators seeking to apply generative AI in a responsible and effective manner. Attendees will gain insight into how to identify and prioritize the right AI opportunities using Lean methods, where AI delivered meaningful productivity gains, where engineering expertise and human review remained essential, and how teams can responsibly incorporate AI into the development of operational and regulatory applications.

The presentation concludes with actionable guidance for adopting generative AI in software development, giving attendees a framework — from opportunity identification and scoring through delivery — for evaluating AI-assisted development opportunities within their own organizations and operational technology environments.