Heidelberg AI News

Video · GitHub ·

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

In short

Levi Strauss redesigned its software development lifecycle around 17 specialized GitHub Copilot agents, with humans retaining approval, prioritization, and accountability. The approach connects requirements, coding, testing, accessibility, performance, security, observability, and deployment through Jira, Confluence, New Relic, Playwright, and MCP, reducing delivery time while enforcing consistent quality gates.

Key points

  • Back Pocket’s new version took 2.5 months versus 9 months for the previous application.
  • The lifecycle was redesigned into planning, building, validation, and release, targeting waiting time rather than merely making people work faster.
  • Agents execute tasks such as testable requirements, code changes, automated tests, security scans, WCAG 2.2 AA checks, performance measurement, and observability setup.
  • Quality gates always run in this order: testing, accessibility, performance, code review, and security.
  • Humans approve specifications and pull requests, while every agent and human action remains versioned, reviewable, and auditable.
  • Levi Strauss reported 51% higher productivity, 54% higher success, and AI-supported pull requests increasing from 26% to 69%.
  • The main lesson is to redesign the development process before adding agents, then introduce reusable agents, skills, MCP servers, and repository templates.

Written by an AI model from the video’s captions and description.

Levi Strauss & Co. built an enterprise multi-agent software development lifecycle on GitHub Copilot's custom agents platform. In this GitHub Universe 2026 session, Jenny Patchen, Sr.

Watch on YouTube The opening of the article at youtube.com.

In this story

A tool or model opens in the Heidelberg AI Curriculum: what it does, what it costs, how to start. A repo opens on GitHub.

Related stories

All stories