The numbers, as Salesforce reports them
Salesforce's own disclosures give a sense of scale: Agentforce annual recurring revenue reportedly reached roughly $800 million, up 169% year-over-year, with more than 29,000 cumulative deals closed since launch. By April 2026, Agentforce agents had produced 734 million "agentic work units," and the average number of unique skills an agent can perform grew from two to six over the course of 2025. Salesforce also reports over $100 million in annualized customer cost savings and a 34% productivity increase across its Agentforce customer base — worth noting these are self-reported, customer-survey-based figures, not independently audited.
Where it's actually working
The pattern across the deployments that hold up isn't "AI does everything" — it's the opposite. The successful early rollouts focus an agent on a single, high-volume task with a predictable structure: order status checks, appointment scheduling, returns initiation, tier-1 billing disputes. One telecom example handles billing disputes, plan changes, and technical troubleshooting without human involvement for tier-1 issues specifically — a narrow, well-bounded scope, not a general-purpose assistant.
Where it's still early
- Data foundation is the actual bottleneck. An agent is only as good as the data it can reason over — which is why so much of the real work in an Agentforce engagement ends up being Data Cloud work first.
- Trust builds slowly, on purpose. Guardrails and human-in-the-loop review aren't a temporary training-wheels phase — they're how adoption survives past the pilot.
- Industry-specific maturity varies. Service and sales use cases have the most public track record right now; Health Cloud-adjacent agentic work is newer, including for us — we'd rather say that plainly than claim delivery experience we don't yet have.
Our take
Agentforce is worth taking seriously, but the entry point matters more than the ambition. Pick the one task your team already does constantly, with a clear definition of "done," and build the data foundation underneath it properly. That's a smaller, more honest starting point than an "AI strategy" — and it's the one that actually ships.
Talk to us about your first Agentforce use case →
Sources referenced for this post
- Futurum Group — "Salesforce Q4 FY 2026 Earnings Show Agentic AI Scaling, Guidance Steadies"
- Salesforce — "Agentic Enterprise Index 2025–2026"
- TechHQ — "Salesforce's Agentforce enterprise bet is paying off and the numbers prove it"
- HyphenX Solutions — "Real-World Agentforce Use Cases That Work in 2026"