Salesforce’s Agentforce enterprise push is working–now comes the harder part

• 
  
  Salesforce closes fiscal 2026 with $800 million in Agentforce ARR.
  
  


• 
  
  Governance gap beneath numbers may be more consequential.
  
  

The headline numbers from Salesforce’s fiscal year 2026 have been released. According to the company’s earnings report, published on February 25, full-year revenue reached US$41.5 billion, up 10% year-on-year. Agentforce ARR hit US$800 million, up 169% year-on-year. The company closed 29,000 Agentforce deals, up 50% quarter-on-quarter in Q4 alone, and delivered 2.4 billion

agentic work

units in Agentforce and Slack.

Salesforce has processed more than 19 trillion tokens to date, up five times year-on-year. For context on what those work units represent: Salesforce defines them as discrete tasks completed by an AI agent; a decision made, a record updated, a workflow triggered. The company introduced this metric with its Q4 results as a deliberate attempt to change the conversation from seats licensed to work actually done.

Marc Benioff, chair and CEO of Salesforce, described the moment in terms that frame the company’s direction: “We’ve rebuilt Salesforce to become the operating system for the Agentic Enterprise, bringing

humans and agents

together on one trusted platform.”

Understanding the claim requires looking at where Agentforce enterprise deployments are actually landing, and where the friction points are emerging.

Salesforce’s relationship with AI in its own operations stretches back further than most enterprises. The company acquired AI startup MetaMind in 2016, launched Einstein the same year as a predictive layer embedded in its CRM, and spent the following years building AI into sales forecasting, pipeline management, and service routing.

That time was characterised by AI that improved human decisions – better signals, faster analysis, cleaner data – but left humans firmly in the execution seat. The change Agentforce represents is different. When it launched Agentforce in September 2024, it was making a public commitment to AI that does not advise but acts; creating records, routing cases, closing workflows, and initiating outreach.

The risk profile of that change is different from predictive AI, and Salesforce acknowledged it at the time with a corresponding focus on guardrails and monitoring. By 2025, Salesforce was describing itself as “customer zero” for Agentforce in its own support functions, handling tens of thousands of customer conversations per week through its own agents.

In the same year, Benioff attributed a reduction of around 4,000 customer support roles to AI agents, describing Agentforce as handling roughly half of all customer interactions.

Deployments and numbers

The customer evidence now sits with the internal story. According to Salesforce’s reported customer outcomes, Williams-Sonoma

deployed

Agentforce in its brand under the name Olive, a customer-facing agent now handling approximately 60% of conversations on the company’s website.

At the US Internal Revenue Service, the Office of the Chief Counsel automated up to 98% of previously manual activities, reducing the time to open a tax court case from 10 days to 30 minutes. A separate IRS division reported saving an estimated 500,000 minutes annually after retiring legacy systems.

Travel platform Engine deployed its Agentforce-powered assistant Eva in 12 days and is projecting US$2 million in savings from reduced human intervention in booking queries. Safari365, an Africa-based tour operator with 35 employees, achieved 62% case resolution through Agentforce after completing a data cleanup that took longer than the agent deployment itself – a detail that carries more editorial weight than the resolution rate.

Across the Agentforce customer base as a whole, Salesforce reports over US$100 million in annualised cost savings and a 34% increase in productivity from agentic and generative AI, based on customer-reported outcomes.

The governance gap

The conditions required to sustain them are where the editorial story gets more complex. According to Salesforce’s own

2026 Connectivity Report

, published in February and drawing on a survey of 1,050 enterprise IT leaders, organisations currently use an average of 12 AI agents, a number projected to climb 67% in two years.

Half of those agents operate in isolated silos not as part of coordinated multi-agent systems. The number of enterprise applications in use grew from 897 to 957 year-on-year, with only 27% of them integrated. As per that report, 86% of IT leaders are concerned that agents will introduce more complexity than value without proper integration.

Primary deployment challenges cited were risk management and compliance at 42%, lack of internal AI expertise at 41%, and legacy infrastructure incompatibility at 37%. A further 96% of organisations reported encountering barriers to using data for AI, with 40% pointing specifically to outdated IT architecture as the bottleneck.

Andrew Comstock, SVP and GM of MuleSoft at Salesforce, said: “The true success of an Agentic Enterprise isn’t found in the sheer number of agents deployed but the overall effectiveness of those agents. We need to think about how they are discovered and orchestrated to work together.”

What Spring ’26 brings

Salesforce’s Spring 2026 release, launched on February 23, reflects a response to deployment realities. According to Salesforce’s release

announcement

, the update introduced Agentforce Builder, a conversational workspace for building and refining agents, with Agentic Enterprise Search, which draws context from 200 external sources and coordinates action in multiple agents.

As per the release notes, the update also brought Agentforce Voice for Financial Services to general availability, letting AI agents to handle banking and collections enquiries. Account Management abilities were added give sales teams an updated view of customer intelligence drawn from CRM data, third-party sources, and conversation history.

The US$45.8 billion question

Salesforce has guided fiscal 2027 revenue at US$45.8 to $46.2 billion an annual 11% growth, with organic revenue expected to re-accelerate in the second half of the year. According to Q4 earnings, more than 60% of Agentforce and Data 360 Q4 bookings came from existing customer expansion, a figure that reflects where the platform is taking root.

The Agentforce enterprise story heading into fiscal 2027 is whether the integration discipline, governance frameworks, and data foundations that make agents reliable at scale can keep pace with the deal count. Safari365’s data cleanup taking longer than the agent deployment is a more fact than the 62% resolution rate that follows it. The enterprises getting the most from Agentforce are the ones that treated data and governance as the precondition.

Salesforce

will have representation at

TechEx North America

, taking place 18 – 19 May 2026 at the San Jose McEnery Convention Centre, California, where Franny Hsiao, Distinguished AI Architect at Salesforce, is among the confirmed speakers.