Salesforce has moved beyond generic AI assistants. On September 11, the company shipped seven named Agentforce agents, each purpose-built for a distinct business function, and it’s backing the launch with a metric that’s hard to ignore: 7 billion Agent Work Units (AWUs) already completed across its customer base.

The agents are Casey, Paige, Carter, Hunter, Marshall, Piper, and Fin. Six are generally available today. The seventh, Hunter, is in pilot with a November general availability date.

Meet the Seven

Here’s what each one actually does:

• Casey handles customer service, fielding inquiries and resolving issues without human handoff
• Paige covers IT and HR, processing internal requests and automating administrative work
• Carter is built for commerce, supporting product discovery, order management, and purchasing workflows
• Hunter manages outbound sales, the only agent still in pilot and the first to use Salesforce’s new long-horizon runtime
• Marshall runs supply chain operations, monitoring inventory and flagging disruptions
• Piper drives inbound pipeline, qualifying leads and routing them through sales processes
• Fin focuses on customer experience, handling interactions across channels at scale

The naming is deliberate. Salesforce isn’t positioning these as tools to configure, but as workforce additions, each with defined job titles and scopes, operating inside a company’s existing permissions, security controls, and business rules.

The Long-Horizon Runtime Is the Real News

Of the seven, Hunter deserves the closest attention, not because of what it does today, but because of how it operates. Hunter runs on Salesforce’s new long-horizon runtime, which pursues goals over weeks rather than responding to a single prompt or completing a single task.

That’s a meaningful architectural shift. Most enterprise AI agents today complete discrete tasks: write this email, summarize that document, answer this question. Hunter is designed to hold a sales pursuit over an extended time window, surfacing the right action at the right moment across a multi-week cycle.

Whether that works at scale in messy real-world environments remains to be seen. But as a signal of where enterprise AI is heading, it’s significant.

The Performance Numbers

Salesforce cited two early adoption figures that show these aren’t just demo capabilities:

Fin is already resolving 79% of Anthropic’s support conversations without a human stepping in. Paige is resolving 70% of administrative requests at Autism Queensland.

These aren’t cherry-picked edge cases from a single flagship client. They’re from organisations with very different customer-facing needs and internal complexity. A 79% autonomous resolution rate in support is commercially meaningful: it means fewer staff hours on routine queries, faster response times, and consistent quality regardless of headcount.

What Business Owners Should Take From This

The pattern here is standardisation. For two years, the enterprise AI story has been about building custom agents, prompt-engineering your own workflows, and standing up bespoke pipelines. That required significant internal capability or expensive consultants.

Salesforce is betting that most businesses want to skip that and deploy something pre-built and job-ready. Seven named agents for seven common business functions is a productised layer on top of what would otherwise be a professional services engagement.

The tradeoff is flexibility. A generic Casey handles customer service the way Salesforce has designed it to. If your support workflows are deeply specific, a pre-built agent may not fit without modification. Businesses that have already invested in custom AI infrastructure may find these agents too constrained.

But for businesses that haven’t started yet and are looking at the complexity of building from scratch, a named agent with documented resolution rates and a clear scope is a more approachable entry point.

What This Means for Business

The 7 billion AWU figure tells you something about where we are in the adoption curve. A year ago, enterprise AI agents were being evaluated. Today, they’re running operational workloads at scale.

The organisations seeing results aren’t the ones still in pilot mode. They’re the ones that committed to specific use cases, gave agents real access to business systems, and measured outcomes. That’s how Fin gets to 79%. That’s how Paige gets to 70%.

If you’re still deciding whether AI agents are real, that window is closing. The companies that moved early are accumulating operational advantages while later movers figure out where to start.

For businesses that want to build rather than buy, the question isn’t whether agents can handle real work. The Salesforce data answers that. The question is whether the right build is a platform lock-in or infrastructure you own.

That’s a strategic decision worth making deliberately, not by default.

---

Enterprise DNA helps businesses deploy AI agents and build the internal capability to operate them. If you’re evaluating how AI agents fit into your operations, start with a discovery conversation.