The Seven New Agentforce Agents Explained

In September 2026 Salesforce launched seven Agentforce agents, each designed for a specific job. Until now, most companies had to design an AI agent from scratch, which meant a lot of configuration before they saw any value. Named agents such as Casey, Piper and Hunter start with the job already defined. This guide explains what each agent does, which are generally available, how the new long-horizon runtime works, and how to choose where to start.

What Are the New Agentforce Agents

An AI agent is software that can understand a request, decide what to do and take action in your systems, rather than only answering a question. Salesforce's new portfolio packages agents for seven common jobs.

Agent	Job	Where it works	Status
Casey	Resolves customer service issues such as FAQs, returns and account changes, with escalation to a person	Voice, SMS, WhatsApp, web chat	Generally available
Fin	Handles complex customer experience workflows	Multiple channels	Generally available
Piper	Engages and qualifies inbound leads	Websites, inboxes	Generally available
Hunter	Works outbound pipeline from research to outreach	Sales systems	Pilot; GA planned November 2026
Paige	Handles IT and HR service requests	Slack, portals, employee tools	Generally available
Marshall	Coordinates supply chain and back-office work	Back-office systems	Generally available
Carter	Helps shoppers find products and check out	In-chat commerce	Generally available

Why Job-Ready Agents Matter for Businesses

The hardest part of an agent project is rarely the AI. It is deciding exactly what the agent should do, what it must never do, and when it hands over to a person. A job-ready agent comes with those decisions largely made, so your team configures and tests rather than designs from zero.

Salesforce has published some early customer results. It reports that Engine's help agent resolves 50% of chat inquiries, that Hunter built 60% of Perk's sales pipeline, that Paige resolves 70% of administrative requests at Autism Queensland, and that Fin resolves 79% of Anthropic's conversations autonomously. These are vendor-reported figures, but they give a sense of the ceiling when data and processes are in good shape.

What Each Agent Does in Practice

Customer service with Casey and Fin

Casey is the obvious first agent for most service teams. It covers the questions that fill queues every day: order status, returns, simple account changes. WhatsApp support matters for businesses in the UAE, the wider GCC and India, where many customers prefer it to email. Fin is aimed at longer, multi-step customer journeys that cross channels.

Sales with Piper and Hunter

Piper responds to inbound leads and qualifies them, which helps when leads arrive outside working hours or faster than your team can reply. Hunter goes the other way. It researches accounts, runs outreach and stays on a deal alongside a human seller.

Internal operations with Paige and Marshall

Paige answers IT and HR requests in Slack and employee portals, such as password resets or leave policy questions. Marshall coordinates supply chain and back-office tasks and keeps an audit record of what it did.

Commerce with Carter

Carter helps shoppers discover products and complete checkout inside a chat conversation. It is most relevant to retailers already using Salesforce Commerce.

How Hunter and the Long-Horizon Runtime Work

Most AI tools complete one task and stop. Hunter is the first agent on Salesforce's long-horizon runtime, which lets an agent pursue a goal across days and weeks. Salesforce describes three parts to it:

• Memory, so the agent keeps context across many sessions.
• Durable execution, so it keeps a plan running over time and adjusts as it learns.
• Dynamic steering, so users can change its behaviour through feedback without rebuilding it.

Our view is that this is where agents become genuinely useful for sales, and also where data quality matters most. An agent that works an account for a month will repeat any mistake in that account's data for a month. Salesforce has said more agents will move onto this runtime over time.

How to Choose Your First Agentforce Agent

1. Start where volume is high and risk is low. Routine service questions or inbound lead qualification are good candidates.
2. Choose a process you already measure. Case deflection, first response time and lead-to-meeting rate are easy to compare before and after.
3. Design the human handoff first. Decide which topics always go to a person and what context the agent passes on.
4. Match the channel to your customers. If most of them message you on WhatsApp, a web chat pilot will not tell you much.
5. Plan for Hunter separately. It is still a pilot. Use the months before general availability to clean account, contact and activity data.

Common Mistakes With a First Agent

• Launching several agents at once. Each one needs its own testing, handoff rules and owner.
• Giving the agent admin-level access. Agents should have their own permission sets with only what the job needs.
• Relying on out-of-date knowledge articles. Casey and Fin can only give answers as good as the content they draw on.
• No named owner. Someone in the business, as well as IT, must be accountable for what the agent does.

How Alu Cloud Consulting Can Help

We help teams choose a first agent, prepare the data and permissions behind it, and run a measured pilot. Before any build, we recommend working through our Agentforce readiness checklist. If you are weighing up the wider set of launches, our Dreamforce 2026 summary puts the agents in context.

Thinking about Casey, Piper or Hunter? Talk to us about a scoped pilot with clear success measures.