September 14, 2026
On September 11, 2026, four days ahead of Dreamforce 2026 in San Francisco (September 15-17), Salesforce announced seven named, "job-ready" Agentforce agents, according to Enterprise DNA. Each has a name and a specific function: Casey handles customer service, Paige covers IT and HR, Carter runs commerce, Hunter does outbound sales, Marshall manages supply chain, Piper works inbound pipeline generation, and Fin handles customer experience. Most of the seven are generally available now. Hunter is the exception, still in pilot with general availability targeted for November 2026, and notable as the first agent to run on what Salesforce calls a long-horizon runtime -- built to pursue goals over weeks rather than operate within a single chat session, per Enterprise DNA's reporting.
Fin's arrival is tied to a separate deal: Salesforce completed its acquisition of the Fin platform on September 10, 2026, one day before folding its capabilities into the new customer-experience agent. Salesforce also introduced the Trusted Enterprise AI Harness, described as a governance layer built for companies already running multiple different AI agent platforms at once. And the company shared usage numbers to back up the scale of all this: Agentforce and Slack have together delivered 7 billion "Agentic Work Units" to date, with 3.2 billion of those in the second quarter alone.
Start with what's genuinely impressive here, because it is: a runtime that can hold a goal across weeks instead of resetting at the end of every session is a meaningfully harder problem than most of what gets marketed as "AI agents" today. Most agent products, ViibeStack's included, are built around discrete tasks with clear start and end points. Hunter's long-horizon design -- tracking an outbound sales motion over an extended window, presumably retaining context, adjusting to responses, and picking back up without a human re-briefing it -- is a real technical step forward if Salesforce can make it work reliably at GA. Don't discount that just because the rest of the announcement is marketing theater.
Here's the part worth sitting with: Casey, Paige, Carter, Hunter, Marshall, Piper, and Fin are not built around any individual company's process. They're templates. Every Salesforce customer buying the outbound sales agent gets the same Hunter, tuned to a generic notion of what outbound sales looks like, not to how your team actually qualifies leads, sequences outreach, or defines a win. Giving it a human name and a job title is a packaging decision, not a product decision -- and it's a smart one, because "pre-hired AI employee" sells a lot better than "configurable workflow agent." That's the shift worth naming: Salesforce is moving Agentforce from assistant framing to headcount framing. But the underlying thing is still a one-size-fits-all persona sitting on top of a subscription you're already paying for, now metered by a new consumption unit most customers have never had to budget for before.
The Trusted Enterprise AI Harness makes the point sharper. It exists to solve a problem: companies running several different AI agent platforms at once need a way to govern them collectively. That's a real problem for a lot of enterprises today. But it's also a problem that mostly gets created by exactly this kind of purchasing pattern -- renting seven separate generic personas, each metered separately, each with its own behavior and blind spots, and then needing a whole additional product just to keep track of what they're all doing. A business that builds one tool shaped around its own process from the start doesn't generate that complexity in the first place, because there's nothing to reconcile across platforms. The governance layer isn't solving a problem inherent to AI agents; it's solving a problem inherent to buying seven of them off a shelf.
If you're already deep in the Salesforce ecosystem, this announcement is worth watching closely, especially Hunter's long-horizon runtime once it reaches GA. But if you're earlier in the process -- comparing no-code app builder options or trying to figure out whether to buy, build, or use something like ViibeStack -- the real question isn't whether an agent has a name. It's whether the tool was built around your workflow or whether you're expected to adapt your workflow to it. A CRM or sales solution that's shaped to your actual pipeline doesn't need a governance harness to manage itself against six other agents, because there's only one system doing the work, priced once, not metered per unit of activity it performs. The Agentic Work Unit numbers Salesforce shared -- 7 billion delivered, 3.2 billion in Q2 alone -- are impressive as a scale claim, but they're also a preview of how billing works once you're on this model: usage-based, stacked on top of your existing license, growing every quarter whether or not the work getting done actually maps to something more valuable than before.
None of this is a knock on Salesforce's execution. Naming your agents and giving them job titles is good marketing, and the long-horizon runtime is a legitimate technical advance. The point is just to look past the names: what's being sold is seven generic personas, a new metering layer, and a governance product to manage the complexity that model creates. That's worth knowing before you sign, whether you're comparing it directly or reading about it on the way to replacing Salesforce with something built around your business instead of theirs.
Sources