Salesforce gave seven AI agents names and job titles

Seven agents instead of one platform

What stands out here is not the number of agents itself, but how they are presented. In its September 11, 2026 announcement, Salesforce introduced seven distinct digital workers with names, job titles and clearly defined responsibilities. Six are already available to customers, while Gunter is still in pilot.

The lineup is organized around familiar business functions rather than technology categories:

• Casey handles customer support;
• Paige processes employee IT and HR requests;
• Carter helps shoppers choose and pay for products;
• Gunter finds prospects and nurtures them over weeks;
• Marshall works with supply operations;
• Piper qualifies incoming leads;
• Finn takes on complex customer cases.

That is where it gets interesting: Salesforce is not selling an abstract platform with a dozen modules, but an understandable role. Not “deploy an automation system,” but “add Casey to support.” It is a meaningful change in the interface between enterprise software and the person deciding whether to adopt it.

Still, the evidence remains thin. According to Salesforce customer claims, Gunter generates 60% of one company’s sales pipeline, while Paige resolves 70% of administrative requests at another. The methodology, sample and comparison conditions are not disclosed, so these figures should be treated as early user claims rather than reproducible performance metrics.

The announcement does not list separate pricing for the seven agents. Agentforce pricing still relies on usage and licensing, which means the “digital employee” framing describes product positioning rather than a literal per-employee payment model.

The real change is in the accountability model

This is a genuine move from assistant to role-based agent, but not yet proof of employee-level autonomy. A Copilot waits for a prompt, whereas Gunter is positioned as an operator able to run a process for weeks. That makes task state, access rights, action controls and reliable handoffs to people critical.

For teams, this framing makes selection easier: a role and expected outcome are easier to understand than a platform feature list. At the same time, it can hide integration complexity. A support agent without access to current data, escalation rules and customer history is only a polished name sitting on top of an old problem.

I would look first at authority boundaries, decision audits and failure behavior rather than anthropomorphic names. If these seven roles hold up in real workflows, the market may genuinely begin comparing AI not with software products, but with staffed business functions. For now, the job titles are more persuasive than the published evidence of autonomy.

We previously covered how AI agent marketplaces are bringing task-specific automation into business workflows. Salesforce’s specialized agents show how that same model is moving into established enterprise platforms.