Analysis
Salesforce introduced a new composable architecture called the Trusted Enterprise AI Harness, built around six capabilities -- Trusted Context, Trusted Agency, Trusted Action, Trusted Governance, Trusted Security and Trusted Models -- alongside a new AI Control Plane, multiple outlets reported. The AI Control Plane is designed to discover, register, govern, observe and control the cost of AI agents across both Salesforce's own products and third-party agent platforms operating inside a customer's environment.
The core problem Salesforce is targeting is real and increasingly common: enterprises are already running multiple agent platforms simultaneously -- Salesforce's own Agentforce alongside competing or complementary agent tooling from other vendors -- without a unified way to see what any given agent can access, what actions it's authorized to take, or what it's actually doing in production. The AI Control Plane's pitch is to sit above that fragmented landscape as a single governance and observability layer rather than requiring IT teams to build custom monitoring for each agent platform separately.
Much of the technology underpinning the Harness isn't new -- Salesforce says it's already available across existing products including Data 360, Informatica, MuleSoft, Agent Fabric, Tableau, Agentforce, Salesforce Guardian and the core Salesforce Platform. What's new is the unified branding and architecture tying those pieces together into a single governance narrative, rather than leaving customers to assemble agent oversight from separately purchased and separately managed components.
The timeline is the detail most likely to get lost in the announcement: the Harness as a complete, generally available product isn't scheduled until early fiscal year 2028, which begins in February 2027 -- roughly a year and a half from this announcement. That's a long runway for a governance framework in a category moving as fast as enterprise AI agents currently are, and it means most of what Salesforce announced this week is a roadmap commitment rather than a product customers can fully deploy today.
Salesforce is far from alone in racing to own the enterprise AI governance layer -- Microsoft's own recent AI code of conduct work and a wave of similar governance announcements from cloud and SaaS vendors this year reflect a broader industry recognition that enterprise customers are demanding oversight tools faster than individual agent products are shipping safety guarantees. Whoever establishes the default governance layer enterprises actually adopt gains meaningful lock-in, since ripping out a governance layer once agent workflows depend on it is a much harder switch than swapping a single point-tool.
A preview announcement with an 18-month runway to general availability is not the same as a shipped product enterprises can rely on today -- Salesforce customers evaluating agent governance needs in the next year will still need interim solutions, and the gap between this week's announcement and FY28 availability leaves real room for competitors to ship comparable governance tooling faster.
Whether Salesforce actually hits its FY28 timeline, and whether the AI Control Plane genuinely works across third-party agent platforms as advertised rather than favoring Salesforce's own Agentforce ecosystem, are the two concrete things enterprise IT buyers evaluating this announcement should be pressing Salesforce on before budgeting around it.
