- Salesforce just launched 7 named Salesforce AI agents — Casey, Paige, Carter, Hunter, Marshall, Piper, and Fin — each pre-built for one specific business function
- Six agents are generally available today; Hunter, the outbound sales agent, is in pilot with full availability planned for November 2026
- Hunter introduces a new “long-horizon runtime” that pursues goals across days or weeks instead of finishing everything in one conversation
- Companies can rename every agent and customize its behavior with Agent Script, Salesforce’s open-source rules language, so end customers may never see the name “Casey” or “Fin” at all
- Salesforce hasn’t disclosed pricing for the new agent portfolio as of launch
Salesforce just gave its AI agents actual names and job titles. On September 11, the company introduced seven new Salesforce AI agents — Casey, Paige, Carter, Hunter, Marshall, Piper, and Fin — each pre-built for one specific business function instead of a general-purpose chatbot. The idea is simple: rather than building an agent from scratch, a company picks the one that already matches the job it needs done.
Meet the Seven Salesforce AI Agents
| Agent | Job | Status |
|---|---|---|
| Casey | Customer service across voice, SMS, WhatsApp, and web chat, with pre-built support for FAQs, returns, and human escalation | GA now |
| Paige | IT and HR requests, resolved across Slack, internal portals, and the tools employees already use | GA now |
| Carter | Shopper assistant that helps compare products and convert with in-chat checkout | GA now |
| Hunter | Outbound sales — researches accounts and works leads over weeks or months | Pilot; GA Nov ’26 |
| Marshall | Supply chain and back-office process automation with a full audit trail of every action | GA now |
| Piper | Inbound lead qualification across websites and inboxes for B2B sales and marketing teams | GA now |
| Fin | Complex customer experience workflows across every channel, powered by custom Fin Apex models | GA now |
Every one of these Salesforce AI agents connects to a company’s existing Customer 360 data, so it works with the customer context and business processes the company already has stored in Salesforce, rather than starting from a blank slate.
Why Names Instead of Feature Lists
Giving each agent a name and job title isn’t just branding. It’s a genuinely useful shortcut for buyers. Startup Fortune put it well: a named agent with a clear job description is easier to justify in a budget meeting than a slide full of feature checkboxes. This mirrors a broader pattern across the industry this year — vendors increasingly package AI capability as a “hire,” not a feature, because it maps more naturally onto how businesses already think about headcount and job functions.
Companies can also rename any of these agents to match their own brand, so customers never actually see “Casey” or “Fin” directly unless the company chooses to keep that name. This matters for enterprise buyers specifically: a company can market its own support experience under its own name while still running on Salesforce’s underlying agent infrastructure.
The Long-Horizon Runtime That Sets Hunter Apart
Hunter is the first agent built on Salesforce’s new long-horizon runtime, which lets it pursue a goal across days or weeks instead of finishing everything in one conversation. So, a seller can ask Hunter to rescue at-risk deals before quarter-end, and Hunter turns that objective into an actual measurable plan rather than a single response.
As the work unfolds, Hunter incorporates new information and adjusts its plan while keeping the seller in control. Three capabilities make this possible:
- Memory — preserves context and progress across sessions, so work doesn’t reset when the conversation ends.
- Durable execution — keeps plans running over time and lets Hunter resume or course-correct as circumstances change.
- Dynamic steering — adapts Hunter’s behavior based on an individual seller’s feedback and direction, rather than following one fixed script.
Salesforce has said more agents across the portfolio will run on this same runtime over time, and that customers will eventually be able to build their own long-horizon agents rather than relying only on Salesforce’s pre-built ones.
What Early Customers Are Actually Reporting
Salesforce backed this launch with specific customer numbers rather than vague claims:
- 50% of Engine’s chat inquiries are fully resolved by its help agent, built on Casey.
- 60% of Perk’s sales pipeline is built by its outbound agent, built on Hunter.
- 70% of Autism Queensland’s administrative requests are resolved by its employee service agent, built on Paige.
- 90% of core shopper journeys at Hibbett are handled by Hibbett AI, which went live in just six weeks.
- Asana reports 4x the conversation volume through its website agent, built on Piper, deployed in 45 days on average.
- 79% of Anthropic’s conversations that reach Fin are resolved without a human stepping in.
These are Salesforce-reported customer figures, not independently verified numbers. They’re the strongest part of Salesforce’s pitch precisely because they’re specific, but treat them as early evidence, not a guarantee every company will see the same results after a quick rollout.
How Companies Can Customize These Agents
Every agent operates inside a company’s existing business rules, permissions, and security setup — Salesforce built the portfolio to sit on top of a company’s current Salesforce configuration rather than requiring a separate system. Beyond simply renaming an agent, companies can use Agent Script, Salesforce’s open-source language for agent behavior, to combine AI reasoning with deterministic rules. This gives administrators granular control over exactly when an agent acts on its own versus when it needs to wait for a human decision — a distinction that matters a lot for regulated industries or high-stakes customer interactions.
Salesforce also rolled out platform-level features alongside the named agents themselves: Multi-Agent Orchestration (already generally available) routes work across multiple specialized agents so they function as one coordinated team on tasks that cross roles or systems. AI Skills in Agentforce Coworker, still in pilot with general availability planned for October 2026, lets an employee teach the platform how to complete a task once, then scales that knowledge across the whole workforce.
Why Salesforce Is Doing This Now
This launch lands just before Dreamforce 2026, Salesforce’s annual flagship event, which is typically where the company makes its biggest product announcements of the year. Salesforce also disclosed that it has delivered 7 billion “Agentic Work Units” across Agentforce and Slack over the past two years, including 3.2 billion in the second quarter alone — a scale claim meant to demonstrate that agentic AI has already moved well past the demo stage for the company’s customer base.
The underlying strategic bet is straightforward: instead of asking every customer to design an agent from a blank canvas, Salesforce pre-packages the most common, highest-value jobs so companies can deploy something useful in days rather than months. Whether that pre-packaged approach actually fits your specific workflow, or whether you’re better off building a custom agent from scratch, is the real evaluation question worth asking before committing.
Frequently Asked Questions
Are all seven Salesforce AI agents available right now?
Six are generally available today: Casey, Paige, Carter, Marshall, Piper, and Fin. Hunter, the outbound sales agent, is currently in pilot, with full availability planned for November 2026.
How much do these agents cost?
Salesforce hasn’t disclosed pricing for the new agent portfolio or its supporting capabilities as of launch. Pricing details typically follow in the weeks after a major Salesforce announcement like this one.
Can a company change the agent’s name and personality?
Yes. Salesforce explicitly designed these agents to be renamed and tailored so they read as an extension of the company’s own brand, not as a generic third-party tool bolted onto their support experience.
Do these agents work with data outside of Salesforce?
They’re built to connect with a company’s existing Customer 360 data first, though Salesforce’s broader Agentforce platform supports integrations with external systems as well, depending on how a company configures it.
What happens if an agent makes a mistake?
Every agent operates inside the business rules and permissions a company sets up, and Marshall specifically includes an audit record of every action it takes. However, the level of human oversight required is something each company configures itself through Agent Script, so the answer varies by deployment.
Final Thoughts
The shift from generic AI chatbots to named, job-specific agents reflects something bigger than a marketing decision. It’s a bet that businesses adopt AI faster when it maps onto roles they already understand, rather than abstract capabilities they have to figure out how to apply. Whether Casey, Hunter, and the rest actually deliver on Salesforce’s early customer numbers at scale is still an open question. But the packaging itself — pre-built, nameable, and ready to deploy in days — is a genuine shift in how enterprise AI gets sold and adopted.
ChatGPT vs Gemini vs Claude: Which One Should You Use in 2026 — for how the underlying models behind agent platforms like this one actually compare.
Read Salesforce’s full announcement for complete technical details on each agent.