Salesforce is making its Agentforce platform much easier to understand: instead of asking businesses to build their own AI agents from scratch, it is now offering ready-made AI workers designed for specific jobs.

The company launched seven named Agentforce agents on 11 September 2026. Six are generally available now, while Hunter, its outbound sales agent, remains in pilot until November.

The idea is straightforward. A business can choose an agent based on the work it wants automated rather than starting with a blank AI system and figuring out how to build the workflow itself.

The new agents cover some of the most common jobs inside sales, customer service and operations:

  • Casey handles customer service across voice, SMS, WhatsApp and web chat. It can answer common questions, process returns and escalate conversations to human employees.
  • Paige handles internal IT and HR requests through Slack, employee portals and existing workplace tools.
  • Carter acts as an ecommerce assistant, helping shoppers compare products and even complete purchases directly through a conversation.
  • Piper handles inbound sales leads from websites and inboxes, including engaging, qualifying and converting prospects.
  • Marshall automates back-office and supply-chain processes while maintaining an audit trail.
  • Fin handles more complex customer-experience workflows across multiple channels.
  • Hunter researches prospects, conducts outreach and works outbound sales pipelines.

The bigger change is agents that keep working

Hunter is particularly interesting because it uses Salesforce’s new long-horizon runtime.

Many current AI agents effectively operate as individual sessions. Give them a task, let them work, and the process eventually ends.

Salesforce wants Agentforce agents to work differently.

Its long-horizon runtime allows an agent to maintain its memory, plan and progress across days or even weeks. The agent can resume work, adjust its approach when something changes and continue moving towards a longer-term objective.

Salesforce has also made multi-agent orchestration generally available. That allows specialized agents to hand tasks to each other rather than forcing one giant agent to handle an entire workflow.

In practice, an inbound sales agent could qualify a prospect before handing the opportunity to another agent or a human salesperson.

Salesforce says Agentforce and Slack have already processed 7 billion “Agentic Work Units”, including 3.2 billion during Q2. It also highlighted customers including Engine, where an AI help agent fully resolves 50% of chat inquiries, and Perk, where an outbound agent has generated 60% of the company’s sales pipeline.

What does Agentforce mean for small businesses?

There is one important limitation: this becomes much more interesting if your business already runs on Salesforce.

A five-person company managing customers through spreadsheets, WhatsApp and a lightweight CRM probably should not migrate to Salesforce simply to get these agents. Agentforce relies on Salesforce’s Customer 360 data, permissions and wider ecosystem, and Salesforce did not publish pricing for the new agents in its announcement.

But for SMBs already using Salesforce, the calculation is very different.

Instead of starting an AI automation project from scratch, you could begin with one repetitive workflow.

For example, a company receiving hundreds of website enquiries could use Piper to respond immediately, ask qualification questions and route promising prospects to sales.

An ecommerce business could use Casey for repetitive “Where is my order?” and return questions while sending unusual cases to a human.

The sensible approach is to start with one workflow where your Salesforce data is already reliable. Measure response times, successful resolutions and how often humans need to intervene before expanding the system.

And keep human approval around expensive or sensitive actions until you know the workflow is reliable.

For agencies, this changes the AI-agent conversation

The opportunity for agencies and consultants may be even clearer.

When a Salesforce client asks for an “AI SDR” or an automated customer-service agent, building another custom chatbot may no longer be the best answer. Salesforce is increasingly providing the agent itself.

The valuable work moves towards integration, data quality, workflow design, permissions, guardrails and reporting.

That is a meaningful shift.

AI agents are gradually moving from experimental assistants businesses have to assemble themselves into packaged software workers designed around recognizable jobs.

Salesforce is betting that businesses would rather hire Piper or Casey than spend months building their own equivalent.