Salesforce has moved AI agents from chat responses to persistent revenue work. At Dreamforce, the company is putting seven job-ready agents into sales, service, commerce, employee support, and back-office operations, including Piper for inbound pipeline generation and Hunter for outbound sales. The bigger change is the new Agentforce runtime: agents can pursue a goal across days and weeks instead of stopping when a conversation ends. For marketing leaders, that is a shift from using AI to produce an asset to using AI to manage a measurable part of the funnel.
The announcement is more than a feature list. Salesforce says Agentforce and Slack have delivered 7 billion Agentic Work Units over the last two years, including 3.2 billion in the second quarter alone. Those are vendor-defined activity measures, not audited revenue outcomes, but they show how quickly AI work is becoming an operating-system layer inside customer data and workflow platforms.
What Salesforce actually launched
Six agents are generally available now. Casey handles customer service across voice, SMS, WhatsApp, and web chat; Paige handles IT and HR requests; Carter assists shoppers and supports in-chat checkout; Marshall orchestrates back-office work with an audit record; Piper engages, qualifies, and converts inbound B2B leads; and Fin manages complex customer-experience workflows. Hunter, the outbound sales agent, is in pilot with general availability planned for November 2026. Salesforce’s launch announcement says the agents connect to Customer 360 and can learn skills, collaborate, and continuously improve.
The runtime behind Hunter adds memory, durable execution, and dynamic steering. In practical terms, a seller can give the agent a goal—such as rescuing at-risk deals before quarter end—and the system can build a plan, select tasks and context, continue working, and request approval where guardrails require it. Multi-Agent Orchestration is generally available, while AI Skills and Agent Optimizer are planned for October.
The proof point marketers should watch
Siemens provides the clearest marketing and revenue example. In a September 15 partnership announcement, Salesforce said Siemens receives more than 2,500 unqualified leads per month for a sales organization of 18,000 people. Two agents now work together: an engagement agent personalizes outreach, and a qualification agent confirms information such as budget and timeline before routing the strongest opportunity to the right seller. Salesforce says Siemens now engages 100% of those inbound leads across 132 countries.
That is not the same as proving 100% conversion or revenue lift. It is a useful illustration of where agents can create leverage: response coverage, qualification consistency, and routing speed. Salesforce also says Asana’s Piper deployment drives four times the conversation volume and averages 45 days to deploy, while Perk attributes 60% of its sales pipeline to Hunter. These are customer-reported claims without published denominators or independent audits, so leaders should treat them as hypotheses to validate, not benchmarks to copy.
“This is a blueprint for the AI era: deep collaboration that goes beyond connecting technology to accelerate transformation, unlock new growth, and create entirely new value for customers,” Salesforce Chair and CEO Marc Benioff said in the Siemens announcement.
Why this changes AI marketing operations
First, the unit of work is moving from campaign to objective. A conventional marketing automation flow executes a fixed sequence. A long-horizon agent can monitor a pipeline condition, revise its plan when data changes, and coordinate with sales or service. That makes lead quality, handoff rules, suppression logic, and approval thresholds more important than a clever prompt.
Second, the CRM is becoming the agent’s operating context. Marketing teams that keep audience definitions, consent, lifecycle stages, product facts, and revenue outcomes in disconnected tools will get inconsistent decisions. The advantage will go to companies with clean first-party data and explicit business rules—not simply the companies with the newest model.
Third, measurement must separate activity from outcome. Track coverage rate, qualified-lead rate, speed to first human touch, meeting rate, pipeline created, pipeline accepted by sales, revenue, margin, opt-outs, and escalation rate. A dashboard that reports agent actions without downstream outcomes is a productivity report, not a growth case.
A 30-day pilot plan for business leaders
Choose one narrow workflow with a clear baseline, such as inbound lead qualification for a single segment. Freeze the decision rights: let the agent research, ask approved questions, summarize, and recommend routing, but require human approval for pricing, claims, unusual discounts, regulated language, or irreversible CRM changes. Create an evaluation set of real historical leads, test false positives and false negatives, and compare agent-assisted performance with the current process.
Then instrument the full funnel. Review transcripts weekly, sample rejected and escalated leads, and publish a scorecard that combines conversion with customer experience and compliance. If the agent improves response coverage but lowers qualified-lead rate, it is not ready to scale. The goal is accountable automation, not maximum autonomy.
Salesforce’s move signals that AI marketing is entering its persistent-operations phase. To design a governed pilot that turns AI search and AI agents into measurable pipeline, contact Real Internet Sales at 803-708-5514 or visit realinternetsales.com.