Every utility technology leader faces the same decision. Run open weight models, closed models like Claude and GPT, or a hyperscaler platform like Microsoft Foundry, AWS Bedrock, or Google Vertex AI. This session cuts through the noise and gives you a working framework you can apply immediately. You will learn how to tier AI use cases by data sensitivity and deployment environment, from customer-facing chatbots to OT-adjacent grid control systems and match each tier to the right model type. We will cover the real tradeoffs: data sovereignty, regulatory exposure under NERC CIP, infrastructure cost, and vendor lock-in. You will also see how to manage model sprawl once five different platforms are running across your organization and leave with a practical decision tool for your own AI roadmap.

Session Takeaways:

• Describe the difference between open weight, closed weight, and hyperscaler-hosted AI models, and where each fits in a utility’s technology stack.
• Design a three-tier framework for matching AI use cases to the right model type based on data sensitivity and deployment environment.
• Assess the regulatory and data sovereignty tradeoffs specific to utilities, including NERC CIP exposure and the OT/IT boundary.
• Analyze the cost and lock-in tradeoffs between self-hosted open weight models and API-based closed models.
• Explain a practical approach for governing AI model sprawl across multiple vendors and platforms.