Salesforce named its Agentforce agents Casey, Hunter and Marshall - what job titles change for automation buyers

What happened

On September 11, Salesforce rolled out seven named AI agents inside Agentforce, each built for one job instead of one chat window. Casey handles customer service across voice, SMS, WhatsApp and web chat. Paige answers IT and HR requests through Slack and internal portals. Carter sells products in a storefront chat, checkout included. Marshall runs back-office supply chain processes with a logged audit trail. Piper works inbound leads from web forms and email into qualified pipeline. Fin handles complex customer experience workflows, built on technology from a prior Salesforce acquisition. Hunter, an outbound sales agent that researches accounts and drafts multi-week outreach sequences, is in pilot now with general availability planned for November.

Six of the seven are generally available today. That is the detail worth noticing: this isn't a demo wave, it's a shipped product line.

What's genuinely new

The individual jobs — support bot, lead qualifier, product-question answerer — aren't new. Vendors have sold versions of each for two years. What changed is the packaging and one piece of infrastructure underneath it.

Salesforce is calling it a "long-horizon runtime": the ability for an agent to hold a goal open across days or weeks rather than resetting after each conversation. Hunter is the agent that actually demonstrates this — tracking a single prospect relationship through research, first outreach, follow-up, and objection handling without losing the thread. The company says this runtime will extend to the other agents later; right now it's real in one of seven.

The other genuine change is Agent Script, a way to write agent behavior as an enforced procedure rather than a loose prompt. That's a direct response to the standard complaint about agentic tools: they wander off-script under ambiguity. Constraining an agent to a defined workflow trades some flexibility for predictability, which is usually the right trade for back-office work.

What it means for a business owner

If you already run on Salesforce, this is the fastest path to a working agent you'll find: the agent already knows your object model, your permission structure, and your existing automation rules, because it lives inside the platform you're already paying for. Turning on Casey or Paige is closer to enabling a feature than commissioning a build. For a company with call volume, ticket backlog, or a slow HR helpdesk, that's a real shortcut.

The catch is pricing, and it is genuinely hard to pin down. Salesforce has run at least three different Agentforce pricing models in the past year — per-conversation ($2), Flex Credits ($0.10 per action in production), and per-user add-ons ($125–$550/user/month) — and the new agents launched without a published price list attached to this announcement specifically. Budget conversations with your account rep before you commit headcount plans around this. Mid-market deployments that stack Agentforce credits with the underlying Data 360 requirement have landed anywhere from $150,000 to $600,000 in year one, according to industry pricing analyses — a number that has nothing to do with what the agent itself costs to run and everything to do with what Salesforce charges to unlock it.

The honest caveat

Named agents with job titles are a marketing move as much as a product one, and it's worth not confusing the two. Casey is still a support bot; giving it a name doesn't change what it can resolve versus what it escalates to a human. The long-horizon runtime — the actual technical advance here — ships in exactly one of the seven agents today, and Salesforce hasn't said when or whether it reaches the rest.

If you're not already deep in the Salesforce ecosystem, this announcement doesn't change your options. These agents don't run outside Salesforce's data model, so evaluating them only makes sense if Salesforce is already where your customer and sales data lives. If it isn't, the cost of moving your data in to get the agent is usually larger than the cost of the agent itself.

What to do about it

If you're a Salesforce customer, ask your rep for the actual per-agent, per-month cost for your volume before you pilot anything — not the sticker price, the number after credits and required add-ons. Pressure-test it against what a comparable narrow-scope build would cost outside the platform. If you're not a Salesforce customer, this isn't your signal to become one; treat it as evidence that "an agent with a job title" is becoming the standard shape vendors sell in, and expect the same framing from other CRM and support platforms within a quarter or two.