17 agents, one SDLC: How Levi Strauss reimagined development on GitHub

by Jenny Patchen

Jenny Patchen, Sr. Director of Shared Platforms & Services at Levi Strauss & Co., walks through the company’s multi-agent development architecture in this GitHub Universe 2026 session. She also discusses reusable agents, team adoption, and lessons from changing the development process.

A multi-agent software development lifecycle

Levi Strauss & Co. built its enterprise workflow on GitHub Copilot’s custom agents platform, using seventeen specialized AI agents across the software development lifecycle. Their responsibilities include:

The architecture connects to Jira, Confluence, and New Relic through Model Context Protocol (MCP), bringing those systems into the multi-agent workflow.

Automated reviews with human approval gates

Every pull request is reviewed by five to eleven independent agents. These agents automatically fix issues and commit changes back to the branch.

Automation does not remove human oversight: the workflow retains strict human approval gates. The session covers how enterprise standards and quality gates fit into this approach.

Reuse, adoption, and lessons learned

Patchen uses the back pocket app’s development lifecycle to explain the process changes, from defining objectives to building the enterprise architecture. She also covers an internal marketplace and reusability, change management, team adoption, and how the company measures impact.

The session presents practical blueprints for organizations considering a multi-agent SDLC, alongside lessons learned from the implementation.

Session chapters