Salesforce pitches a "Trusted Enterprise AI Harness" as the fix for running 3+ agent platforms at once

Salesforce announced the Trusted Enterprise AI Harness on September 10, 2026, ahead of its Dreamforce conference. It's a composable architecture bundling six capabilities (Trusted Context, Trusted Agency, Trusted Action, Trusted Governance, Trusted Security, and Trusted Models) under a new "AI Control Plane" meant to manage agents and models across an organization regardless of which vendor built them. Many of the underlying components are available now; the unified experience and newer capabilities roll out in Salesforce's early fiscal FY28.

The fragmentation number behind the pitch

Salesforce is framing this around a specific data point: a VentureBeat Intelligence survey found 85% of enterprises already run two or more agent orchestration platforms, averaging 3.1 per company. President and Chief Platform & Engineering Officer Rohan Kumar put the positioning directly: "The Agentic Enterprise won't be defined by which model a company chooses." Salesforce is positioning the Harness as the governance layer that sits above whichever models a company already has scattered across its stack, rather than competing on models.

A customer on lock-in

Rocket Mortgage CTO Shawn Malhotra, quoted in Salesforce's own announcement, said "we don't want to bet our future on one closed stack" -- a named enterprise customer publicly stating the vendor-lock-in concern that every multi-agent-platform buying decision is implicitly about. For engineering and finance leaders already juggling multiple agent platforms and the spend that comes with each, the 3.1-platforms-per-company figure is itself a useful benchmark, independent of whether Salesforce's specific answer to it wins out.