Salesforce Hires Seven New Employees Named Casey, Paige, Carter, Hunter, Marshall, Piper, and Fin
None of them are people. All of them have job titles, performance metrics, and in Hunter's case, a multi-week memory — which is more career development than most of Salesforce's actual sales reps get.
On September 11, Salesforce introduced seven new “job-ready” AI agents, each with a first name, a defined role, and — the company was careful to specify — a place on an actual org chart. Casey handles customer service across voice, SMS, WhatsApp, and web chat. Paige fields internal IT and HR requests. Carter helps shoppers pick things out and checks them out. Marshall runs supply-chain back-office work with an audit trail. Piper qualifies inbound sales leads. Fin resolves complex customer-experience workflows, built on technology from an acquisition Salesforce made earlier this year for $3.6 billion. And Hunter — currently in pilot, general availability slated for November — works an outbound sales pipeline from research through outreach, over weeks at a time, which Salesforce is functionally introducing as a colleague rather than shipping as a feature.
This is, by any reasonable reading, a hiring announcement for software. Six of the seven are already “generally available,” a phrase that in any other context means “you can buy this” and here also apparently means “please welcome them to the team.”
The headline capability is what Salesforce calls a “long-horizon runtime,” debuting with Hunter. Rather than resetting after each task, Hunter is built to “develop long-term work plans and reuse data across chat sessions” — to remember what it was doing last Tuesday and pick the thread back up, the way a human salesperson nurtures a lead over a quarter instead of forgetting them at 5pm. Salesforce is marketing persistent memory and durable, multi-week task execution as the frontier innovation here, which means the frontier of enterprise AI in 2026 is an agent that can remember what it was doing yesterday — a bar so low it should be an OSHA violation for the industry that cleared it.
The performance numbers Salesforce is citing read like employee-of-the-month plaques: one customer had 50% of inquiries fully resolved by an agent, another had 60% of its sales pipeline built automatically, a third saw 90% of shopper journeys handled with no human involved. Buried in the list of case studies is a detail worth sitting with for a moment: Fin, the customer-service agent, is reportedly resolving 79% of support conversations autonomously for Anthropic — the AI safety company whose CEO spent this same week publishing an essay about the dangers of AI systems operating with too much autonomy, too fast, without enough human oversight. Anthropic outsourced the bulk of its own customer support to exactly that.
None of this is subtle marketing. Salesforce isn’t calling these “workflows” or “automations” or even “copilots” — the vocabulary the rest of the industry has spent two years settling into specifically to avoid the implication that software is now doing headcount’s job. Salesforce is doing the opposite: first names, job functions, a “job-ready” framing lifted straight from a staffing agency’s landing page, and a roadmap in which the newest hire (Hunter) starts as an intern (pilot) before earning full-time status (GA) in November. It is, structurally, an onboarding calendar for a person who does not exist and never sleeps.
There’s an obvious, mundane logic to this. “Deploy our new autonomous supply-chain automation suite” is a sentence that requires a procurement meeting. “Meet Marshall” requires nothing but an introduction. Giving software a name and a job title collapses the psychological distance between buying a tool and hiring a worker, and Salesforce, which sells enterprise software for a living, understands better than almost anyone that the second framing closes deals the first one doesn’t.
Developers who want to keep Casey, Paige, and the rest from improvising can write their behavior in “Agent Script,” a constrained syntax Salesforce built specifically to reduce hallucination by forcing agents to follow prescribed task-execution paths rather than reasoning freely — which is to say, the new hires come with an employee handbook so restrictive it’s technically a programming language. An updated “Agentforce Coworker” platform, meanwhile, adds multi-agent orchestration, so Casey, Paige, Carter, Hunter, Marshall, Piper, and Fin can now hand work off to each other and to specialized sub-agents beneath them, forming what is, functionally, a small AI-run department with no meetings, no PTO requests, and — per the marketing copy, at least — no need for a manager.
That last point lands with unfortunate timing. Just one day before Salesforce introduced its seven fully autonomous, memory-equipped, sub-agent-commanding new hires, Meta was reportedly asking its own human employees, reassigned into its Applied AI division after a round of flattening that eliminated thousands of management jobs, whether they’d like to become managers again. Somewhere between those two announcements is the actual state of enterprise AI in September 2026: the software is ready to run itself. The humans supervising it are being rehired to make sure it doesn’t.
Sources: