Salesforce data: AI generates 20% of production code, not Anthropic’s predicted 90%

Salesforce’s AI coding tool reveals the real-world impact of AI on software development, contradicting predictions of immediate developer displacement. While Anthropic’s CEO predicted AI would write 90% of code within months, Salesforce‘s actual data shows Agentforce generating 20% of production-level APEX code. This gap between prediction and reality highlights how AI is transforming development roles without replacing humans, offering valuable insights into how AI coding assistants are actually being used at enterprise scale.

The big picture: Salesforce’s Agentforce coding assistant demonstrates significant but measured AI adoption in enterprise development, with 35,000 monthly users and 10 million lines of accepted code.

• Agentforce is generating approximately 20% of all APEX code deployed at Salesforce in the past month, far below the 90% prediction made by Anthropic CEO Dario Amodei.
• The tool has saved an estimated 30,000 developer hours monthly, showing meaningful productivity gains without triggering massive workforce reductions.

What they’re saying: Salesforce executives believe AI will handle most initial code creation while developers evolve into refinement specialists.

• “The vast majority of development — at least what I call the first draft of code — will be written by AI,” explained Jayesh Govindarajan, Senior Vice President of Salesforce AI.
• “But what developers do with that first draft has fundamentally changed.”

Behind the numbers: Salesforce’s data provides a rare glimpse into enterprise-scale AI coding adoption based on verified production deployments rather than controlled experiments.

• Agentforce’s 20% contribution to production code represents actual deployed software, not merely suggestions or experiments.
• The tool is specifically optimized for Salesforce’s APEX language, which may contribute to its successful adoption compared to general-purpose coding assistants.

The new development workflow: AI coding assistants are transforming developers into specialized reviewers, editors, and architectural decision-makers.

• Developers increasingly spend time refining AI-generated code, focusing on integration, security, and edge cases rather than writing initial implementations.
• The most productive developers have mastered the art of “vibe coding” — describing their intentions conversationally to AI tools rather than using rigid programming syntax.

Why this matters: The gap between AI hype and reality provides valuable lessons about technological adoption patterns and the evolving nature of knowledge work.

• Salesforce’s data suggests AI coding assistants enhance developer productivity without the dramatic workforce displacement many feared.
• This pattern of augmentation rather than replacement mirrors historical technological transitions dating back to the industrial revolution.

What to watch: The evolution of developer roles will likely accelerate as AI coding tools improve, potentially creating new job categories.

• Companies that invest in retraining developers for AI collaboration may gain significant competitive advantages.
• The gap between AI-generated first drafts and production-ready code represents a key opportunity for human expertise and judgment.