Salesforce just handed its AI agents job titles. On September 11, 2026, the company announced seven named agents, Casey, Paige, Carter, Hunter, Marshall, Piper, and Fin, each built for a specific role in sales, service, commerce, or the back office. This is the clearest sign yet that the AI agent workforce concept is moving from buzzword to product line. Over the past two years, Salesforce says its agents completed 7 billion Agentic Work Units across Agentforce and Slack, with 3.2 billion of those in the second quarter alone. That volume points to a real shift in how companies staff repetitive, high-volume work.

This matters for anyone evaluating AI agents right now, not just Salesforce customers. The pattern Salesforce is following, giving agents names, defined jobs, and the ability to work over days or weeks instead of a single chat, is likely to show up across every major platform. Here is what the launch actually includes, what the data shows, and how to think about building or buying an AI agent workforce of your own.

What the New Named AI Agents Actually Do

Each of the seven agents targets a specific job rather than a general-purpose chat interface. Casey handles customer service across voice, SMS, WhatsApp, and web chat, with prebuilt flows for FAQs, returns, and account management. Paige resolves IT and HR requests inside Slack and employee portals. Carter helps online shoppers compare products and checks them out in-chat. Piper works inboxes and websites to qualify inbound leads for B2B sales teams. Marshall orchestrates back-office processes end to end and keeps an audit record of every action it takes. Fin, built on the Intercom acquisition, resolves complex support workflows across channels.

Six of the seven agents are generally available now. Hunter, the outbound sales agent, is still in pilot, with general availability set for November 2026. Companies can rename any of these agents and tune them to their own brand voice, so the names Casey or Fin are really templates rather than fixed identities. The broader idea is that businesses no longer need to build an agent from a blank prompt. They start with a role that already has the skills, actions, and data connections that job requires, then customize from there.

Inside the Long-Horizon Runtime Powering Hunter

The more consequential piece of this launch is technical, not cosmetic. Salesforce built a new long-horizon runtime specifically so agents can pursue a goal across days or weeks instead of finishing when a chat window closes. Hunter is the first agent running on it. A seller can ask Hunter to rescue at-risk deals before quarter end, and Hunter turns that into a plan, decides what tasks to run, and keeps working as new information arrives, checking back in when a decision needs a human.

Three capabilities make this possible. Memory preserves context and progress across sessions, so an agent does not restart from zero every time someone opens a new conversation. For a closer look at how AI agent memory actually works under the hood, the architecture behind this is worth understanding before you evaluate any long-horizon platform. Durable execution keeps a plan running over time and lets the agent resume or adjust course as circumstances change. Dynamic steering adapts the agent’s behavior based on direct user feedback, similar to how a manager might redirect a team member mid-project.

Early results back up the pitch. Salesforce reports that 60 percent of Perk’s sales pipeline is now built by Hunter, 70 percent of Autism Queensland’s administrative requests are resolved by its Paige-based agent, and 90 percent of Hibbett’s core shopper journeys are handled by its commerce agent, which went live in six weeks. Asana’s inbound agent, built on Piper, now drives four times the conversation volume it had before.

How to Apply This If You Are Not a Salesforce Shop

You do not need Agentforce specifically to learn from this launch, and the same lessons apply whether you are scaling AI agent fleets from pilot to production on a different platform or starting from scratch. The practical lesson is that AI agent workforce planning increasingly starts with a job description, not a prompt. Before adopting any agent platform, map the actual role you want automated: what triggers the work, what systems it touches, what a human needs to approve, and what “done” looks like. That mirrors how Marshall’s audit record works, logging every action so a person can trace what the agent did and why.

Long-horizon work also changes how you should measure success. A single-session chatbot gets judged on resolution rate for one conversation. An agent working a goal over weeks needs milestone tracking instead, similar to how a sales manager would check pipeline progress rather than grading one call. If you are piloting an agent for outbound sales, account research, or back-office reconciliation, build in checkpoints where a human reviews the plan before the agent continues, not just after it finishes.

Finally, treat agent memory and permissions as separate design decisions. An agent that remembers context across sessions is more useful, but it also needs clearer boundaries on what it can act on without asking first. That balance, more autonomy paired with tighter guardrails, is the real theme behind Salesforce’s Agent Script feature, which lets teams combine AI reasoning with deterministic rules for how an agent is allowed to act.

The Gap Between Naming Agents and Trusting Them

Giving an agent a name and a job title makes it easier to sell, but it does not automatically make it easier to govern. Forrester’s 2026 research on agentic AI adoption found that most enterprises are still chasing scaled deployment rather than catching it, with only a small share of companies running true multi-agent systems in production beyond early pilots. That gap is exactly why choosing the right AI agent governance platform matters as much as picking the agent itself. Naming an agent Hunter or Casey can also blur accountability. When a named agent makes a mistake on a live sales call or a customer refund, the postmortem still needs to trace back to a specific rule, permission, or training gap, not just a friendly persona.

Expect more vendors to follow Salesforce’s lead by branding agents as job-ready hires rather than generic assistants. The businesses that benefit most will be the ones that treat the name as an interface choice and keep investing in the unglamorous parts: audit trails, approval gates, and clear ownership when something goes wrong.

Key Takeaways and What to Watch Next

Salesforce’s named agent lineup shows where the AI agent workforce is headed: role-specific agents with memory, long-horizon goal pursuit, and built-in audit trails, not one general chatbot trying to do everything. Early customer results, from Hunter’s pipeline contributions to Hibbett’s six-week rollout, suggest this approach can move faster than building agents from scratch. At the same time, naming an agent does not replace the governance work of defining permissions, checkpoints, and accountability before it touches real customers or revenue.

If you are exploring AI agents for your own business, start by defining the job before you pick the tool. For more breakdowns of the latest AI agent platforms, deployment frameworks, and tools, explore the resources at BigAIAgent.tech. What would you name your first AI agent, and what job would you actually trust it to finish without checking in?