On 11 September 2026, Salesforce gave its AI agents names, job titles and, for one of them, a runtime that lets it keep working on a goal for months. The announcement bundled seven Agentforce agents — Casey, Paige, Carter, Hunter, Marshall, Piper and Fin — each built for one job function, alongside a new "long-horizon" runtime that is the more substantial change buried under the branding. A day earlier, on 10 September, Salesforce completed its acquisition of Fin, the AI customer-agent business it bought from Intercom, folding a 30,000-customer support product directly into the lineup as the seventh name.
Seven jobs, seven names
Each agent is scoped to a specific function rather than sold as one general-purpose assistant: Casey handles customer service across voice, SMS, WhatsApp and web chat; Paige resolves IT and HR requests through Slack and internal portals; Carter helps shoppers discover, compare and check out products in-chat; Hunter works a sales pipeline from research through outreach; Marshall orchestrates back-office processes with deterministic execution and a full audit trail; Piper engages and qualifies inbound leads for B2B sales and marketing; and Fin, the newest addition via the Intercom acquisition, handles complex customer-experience workflows across every channel, with an average resolution rate Salesforce and Fin's own materials put at 76% across its existing base of more than 30,000 customers.
None of this is a new capability in the sense of new AI models or new reasoning ability. It is a packaging decision: instead of one configurable "Agentforce agent" that a buyer has to scope and name themselves, Salesforce is shipping seven pre-scoped personas with job titles attached.
The part that actually changes what these agents can do
The more consequential piece of the announcement is a new long-horizon runtime, which lets an agent pursue a goal across days or weeks instead of finishing within a single session. It has three named components: memory that persists context between sessions, spanning months rather than resetting each conversation; durable execution that keeps a multi-step plan running in the background; and dynamic steering that lets a person adjust the agent's behaviour mid-task without restarting it. Hunter is the first agent running on this architecture, currently in pilot, with general availability scheduled for November 2026. Salesforce says more agents will move onto the long-horizon runtime over time, and that developers will eventually be able to build their own long-horizon agents on Agentforce — that capability is not available yet, so treat it as a stated roadmap item rather than something to plan around today.
Why the naming choice is worth studying, not dismissing
It is easy to read seven personified agent names as marketing gloss on top of an underlying platform. It is also a legible answer to a real problem every team building an AI feature runs into: buyers and internal stakeholders struggle to reason about what "an AI agent" can and can't do when it is described only in terms of capability. A name and a job title compress a large amount of scope information into something a non-technical buyer, or a colleague evaluating whether to trust it with a task, can hold in their head immediately. "Hunter handles sales pipeline research and outreach" is a claim a sales manager can staff around, delegate to, and hold accountable, in a way that "an autonomous AI agent with sales capabilities" is not.
That is worth borrowing directly if you are shipping an AI feature inside your own product. Rather than describing your AI functionality only by what model or capability powers it, consider whether giving a specific, scoped piece of it a name and an explicit job description would make it easier for a customer to understand exactly what it will and won't do — and easier for your own team to talk about its limits without hedging every sentence with "the AI."
What to actually watch
If you compete with any part of the customer-support-AI category, the more immediate fact than the branding is that Salesforce just acquired a 30,000-customer, 76%-resolution-rate competitor and folded it into its own platform with distribution to Salesforce's existing customer base. That is a bigger competitive shift than the seven names attached to it. And if you build or plan to build agents that need to work across days rather than single sessions, the memory-plus-durable-execution-plus-steering pattern Salesforce just shipped for Hunter is a concrete reference architecture worth studying now, months before Salesforce's own developer access to it exists.
The bottom line
Salesforce's seven named Agentforce agents are mostly a packaging decision, and a reasonable one: specific names and job titles make AI scope easier for a buyer to reason about than a generic capability list. The substance is underneath it — a long-horizon runtime giving Hunter persistent, months-long memory and durable multi-day execution, reaching general availability in November, plus a freshly absorbed 30,000-customer support-AI competitor in Fin. If you're pricing or positioning your own AI features, both are worth studying before your buyers start asking why your AI doesn't have a name.