Salesforce’s second Agentic Enterprise Index, published this week, shows AI agent deployment inside businesses growing far faster than headcount debates about AI usually assume. Drawing on usage data from companies running Agentforce alongside a survey of thousands of respondents, the report finds that the number of agents activated per organization has nearly tripled since February 2025, while the average time to create and deploy a new agent dropped 53% to under two days. Agent activity itself is compounding: total output from these systems, measured in what Salesforce calls Agentic Work Units, grew at a 15% monthly rate as of April 2026.
Why it matters for the marketing leader: this is not evenly spread. Retail carries the highest share of agentic work output at 22% of the total, but public sector agent output grew 227-fold over the same period, the fastest expansion of any sector Salesforce tracked. That gap matters for budget planning, because it means the ceiling on agent adoption is not technology readiness, it is how quickly a given function can define the workflows worth automating. “Whether you’re spinning up agents to operate at massive scale or orchestrating them through deep, multistep pipelines, the bottom line is they’re shipping real value,” said Joe Inzerillo, Salesforce’s President of Enterprise AI and Technology.
The original angle worth tracking: deployment speed, not agent count, is becoming the real adoption metric. A team that can stand up a working agent in under two days can iterate on failed use cases fast enough to make the occasional bad agent cheap, which changes the risk calculus around experimenting with agentic workflows at all. That mirrors the caution seen elsewhere in the market, where marketers still trust AI advice more than they trust AI acting unsupervised, even as deployment friction keeps falling and use cases like agent-driven fan engagement move from pilot to production.
Source: Salesforce
