From the source
- Salesforce scaled Claude Code to thousands of engineers and doubled effective output, with fewer bugs per line of code than before, its engineering leaders say.
- The company measures AI coding ROI with an "effective output" score it built with Stanford researchers.
- Leadership describes an agent maturity curve that moved engineers from simple prompts to autonomous agents across the software development lifecycle.
- A decades-old codebase was made headless in weeks instead of a year, and a marketplace exists for sharing agent skills.
- Incident agents find the right team automatically, and Claude Opus tracked down a bug in 24 hours after human attempts failed.
- Salesforce's guardrails produced double the code volume with fewer bugs, and the team replays benchmarks to test new models.






