Beyond the Prompt: Governing AI Behavior Like Software Releases

AI behavior can change when a prompt, model, threshold, or surrounding context changes—even when the application code does not. This ELC roundtable asks engineering leaders to govern those changes with the same seriousness applied to software releases.

Why this event matters

Many teams still treat prompts and model settings as configuration details. In production, they can alter quality, safety, cost, and customer experience. Release discipline creates a path to test changes gradually, observe their effects, and reverse them before a small issue becomes a broad incident.

What the conversation will cover

• Progressive rollout for changes to prompts and AI configuration
• Experimentation that measures behavior rather than only deployment success
• Kill switches and rollback paths for production AI features
• Governance through the delivery pipeline
• A live look at Harness AI Configs

Who should attend

The online session is built for engineering executives, platform and AI leaders, product-minded technical managers, and teams responsible for shipping AI features into production.

The DevCuration view

The practical shift is to stop treating model behavior as something outside the release process. If a change can affect users, cost, or risk, it needs ownership, observability, controlled exposure, and a tested way back.

Event details

• Date: 2026-08-05
• Time: 10:00 AM–11:00 AM PT
• Format: Online
• Location: Online
• Hosted by: Engineering Leadership Community, Harness
• Featured speakers: Aaron Newcomb, Harness; Nicolás Zelaya, Harness
• Registration: Official event page

Registration requirements, capacity, and access details can change. Use the official registration page above for the latest information from the organizer.