Salesforce job-ready agents: the runtime is the news
Salesforce shipped seven named Agentforce agents on September 11 plus a long-horizon runtime. The durable execution layer matters more than the names.
Salesforce gave its AI agents first names. Casey, Paige, Carter, Hunter, Marshall, Piper, Fin — a roster of job-ready agents announced September 11, four days before Dreamforce. The names are the part everyone quoted. They are the least useful thing in the announcement. The line worth reading sits underneath the roster: a long-horizon runtime that lets one agent chase one goal across days and weeks instead of one chat window.
What actually happened
Per Salesforce's announcement, seven agents ship with assigned job functions and customer-renameable names. Casey handles service across voice, SMS, WhatsApp and web chat. Paige covers IT and HR requests in Slack and portals. Carter is the shopper agent — discovery, comparison, checkout. Piper qualifies inbound leads. Marshall orchestrates back-office supply chain work. Fin handles complex customer-experience workflows. Six are generally available now. Hunter, the outbound sales agent, is in pilot with GA in November 2026.
The runtime is the actual engineering claim. Salesforce describes three properties: memory that survives across sessions, durable execution so a long-running job resumes instead of dying, and dynamic steering so a human can course-correct an agent mid-goal without restarting it.
Note what is not in the announcement: pricing. And note the calendar — the Trusted Enterprise AI Harness that governs all of this ships in FY28, which starts February 2027. Agents now, control plane later.
Why job-ready agents matter for your business
A named agent is a headcount metaphor, and the metaphor is the pitch. Do not price it that way. "Carter" is not a merchandiser you hired; it is a workflow with a face on it, and you will still own the catalog data, the returns policy, and the escalation path it runs on.
The durable execution part is real and you should steal it. An agent that survives a process restart, keeps state between steps, and can be redirected mid-run is not magic — it is a job queue with persisted state and retries. Temporal, Inngest, or a Postgres table with a status column and an idempotency key will get a three-person shop most of the way there. The reason Salesforce is shipping it as a headline feature is that most "AI agents" in production today are a single prompt in a request handler that loses everything when the request ends. That is the gap, and it is a plumbing gap.
If you already run Commerce Cloud or Service Cloud, Carter and Casey are a short evaluation. If you do not, the buy-in is a platform, not a feature.
Key takeaways
- Salesforce announced seven named Agentforce agents on September 11; six are GA, Hunter goes GA November 2026
- The long-horizon runtime — persistent memory, durable execution, mid-run steering — is the substantive change
- No pricing was published in the announcement; the governing control plane does not ship until FY28
- Durable execution is a solved plumbing problem: a queue, persisted state, retries, idempotency keys
- Most agents that fail in production fail because they hold no state past one request, not because the model was weak
- A named agent is a metaphor, not headcount — evaluate it as a workflow with a vendor attached
Long-running agents need a runtime, not a bigger prompt. We build agent workflows on durable job infrastructure you own — state in your database, retries you can read, a kill switch that works. See how we build AI agent systems, or tell us which process should run unattended.
Sources: Salesforce: Agentforce job-ready AI agents, Salesforce: Trusted Enterprise AI Harness.
- #salesforce
- #ai-agents
- #agentforce
- #durable-execution
- #automation
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
Senate AI duty of care draft: model releases get a veto
Thune, Cruz and Klobuchar are drafting a federal AI duty of care with power to block unsafe model releases and preempt state law. What it means for your stack.
Read itOpenAI won't IPO in 2026: your core vendor stays private
Sam Altman says an OpenAI IPO in 2026 would be 'ill-advised' given safety concerns. What a private core AI vendor means for the small businesses built on its API.
Read it