Salesforce Plans Trusted Enterprise AI Harness to Improve Agent Task Success
Salesforce plans to launch its Trusted Enterprise AI Harness, software code and instructions meant to ground generative AI outputs in shared business context and steer AI agents toward tasks more predictably. The preview arrives at Dreamforce in San Francisco (Sept. 15–17) with broader availability in early next year. Salesforce cites benchmarks where tuning a harness around a smaller model raised task success from 29.2% to 78.0%, without changing model weights.
Coverage (5)
- Salesforce introduces new AI agents to automate sales, support tasks (siliconangle.com)
- Salesforce introduces Enterprise AI Harness, AI Control Plane (siliconangle.com)
- Salesforce Introduces the Trusted Enterprise AI Harness (salesforce.com)
- Many models, many agents, many tasks: Salesforce’s new Enterprise AI Harness seeks to ground all in your shared business context (venturebeat.com)
- Companies already run 3 agent platforms. Salesforce's new Enterprise AI Harness wants to govern all of them. (venturebeat.com)