The developer as conductor: Leading an orchestra of AI agents with the feature flag baton

Dynatrace executives say feature flags are AI's safety net

Dynatrace executives say feature flagging is now an essential "safety net" for managing AI-generated code. The comments follow the observability giant's recent acquisition of feature flagging startup DevCycle.

Michael Beemer, a senior product manager at Dynatrace, and Andrew Norris, co-founder of DevCycle and now a principal product manager at Dynatrace, discussed the integration on The New Stack Makers podcast. They argued the merger creates a "360-degree view" of software performance at the feature level.

Solving the visibility gap in AI development

The push is driven by the rise of "agentic development," where AI agents generate vast amounts of code rapidly. Norris stated this fast-paced output is already creating bugs and problems.

Feature flags, he explained, allow teams to deploy this AI-generated code in a controlled way. They enable a "human in the loop" to approve changes in live environments before a full release, de-risking enterprise adoption of AI tools.

"It’s really de-risking and accelerating the adoption at an enterprise software level of using these agenetic development tools that are augmenting teams," Norris said.

The developer becomes an orchestra conductor

Norris predicted a fundamental shift in a developer's role, borrowing a metaphor from Dynatrace's Alois Reitbauer. "The real role the developer is going to be a conductor — conducting an orchestra of agents," he said.

In this model, feature flags act as the developer's baton. They provide the precise control needed to orchestrate multiple AI agents without sacrificing system stability or quality.

The goal is to let developers embrace the speed of AI-assisted coding while maintaining oversight and the ability to quickly roll back changes.

Open standards and platform integration

A key part of Dynatrace's strategy involves industry standards to prevent vendor lock-in. Beemer highlighted the importance of the Cloud Native Computing Foundation's Open Feature project.

"We wanted to try to standardize how developers worked with feature flags in their code, regardless of the back end," Beemer said. The native integration of DevCycle into the Dynatrace platform aims to provide developers with a unified workflow and complete observability into feature performance.

The combined platform is designed to give teams critical insights, including:

• Real-time performance impact of newly launched features.
• User experience data tied directly to specific feature flags.
• The ability to correlate system errors with recent feature deployments.

As AI continues to reshape software development, Dynatrace is betting that observability and controlled deployment will become non-negotiable requirements for enterprise teams.