Salesforce Cracks Code on Scaling Agentic Coding Across Thousands of Engineers
Salesforce Engineering's Agentic Shift Hits Milestones
The company has scaled its agentic coding to 15,000 engineers, resulting in significant productivity gains. In the past year, work items completed per developer have increased by 90.5%, pull requests merged per developer are up 88.1%, and effective output is up 200.3%. The numbers demonstrate the success of Salesforce's approach to adopting AI tools.
The company attributes its success to a culture that encourages engineers to identify gaps in technology and build solutions themselves, rather than relying on vendors or top-down curation. This approach has allowed teams across the organization to experiment with different approaches and share their innovations with others.
A key part of Salesforce's strategy was the Agent Coding Maturity Curve, a shared language for understanding where engineers are in their adoption of AI tools. The curve helps managers coach engineers and identify specific behaviors that separate stage 3 from stage 6.