The bear case against Salesforce has always been seductive: autonomous AI agents replace the human workers who need CRM seats, so the more powerful agents become, the smaller Salesforce’s addressable market. Dreamforce 2026, opening tomorrow in San Francisco, is where that argument gets its most direct rebuttal yet.
On September 11, Salesforce announced seven named “job-ready” Agentforce AI agents, each purpose-built for a specific business function. They are Casey for customer service, Paige for IT and HR, Carter for commerce, Hunter for outbound sales, Marshall for supply chain, Piper for inbound pipeline, and Fin for customer experience. Six are generally available now, while Hunter remains in pilot and uses a long-horizon runtime that can pursue goals over days and weeks instead of a single chat session.
Early customer evidence is harder to dismiss than most vendor claims. Salesforce says Perk is now building 60% of its North America SDR-generated pipeline through its prospecting agent. Over the past two years, Salesforce has delivered 7 billion Agentic Work Units across Agentforce and Slack, including 3.2 billion in its fiscal Q2. Those are not press-release metrics. They are billable events, and they suggest the product is actually running in production.
The smarter question, though, is what Salesforce charges for next.
Agentforce pricing now comes in three models: $2 per conversation, Flex Credits at $500 per 100,000 credits, or per-user licensing starting at $125 per user per month. Having three live pricing models at once is unusual, and it reflects genuine difficulty translating agentic AI usage into a bill customers can predict in advance. That uncertainty cuts both ways. For buyers, it creates implementation risk. For Salesforce investors, it raises a legitimate question about revenue predictability as the mix shifts away from seat-based subscriptions.
The Governance Layer Is the Longer Game
The agents are the headline. The governance layer is the franchise.
Alongside the agent portfolio, Salesforce introduced the Trusted Enterprise AI Harness, a six-part governance framework for enterprises already running agents from multiple vendors, covering Trusted Context, Trusted Agency, Trusted Action, Trusted Governance, Trusted Security, and Trusted Models, plus a new AI Control Plane for managing agents and AI systems across an entire organization.
The strategic logic behind the Harness rests on a documented reality: 85% of enterprises run two or more agent orchestration platforms simultaneously, with an average of 3.1 platforms per company. This is a strategic move to position Salesforce as the central management layer for agentic AI across the entire enterprise, not just within its own ecosystem. That ambition matters more than any individual agent, because it makes Salesforce’s value proposition independent of which underlying AI model wins. Whether customers run Claude, GPT, or Gemini underneath, the governance layer still needs to sit on top.
Because the capabilities live in a shared harness rather than inside individual agents, Salesforce says they become reusable across the enterprise, with context shared across agents and models, actions and workflows securely invoked wherever needed, and governance applied consistently as AI moves across the business. Broad customer availability is expected in early 2027. Packaging and pricing have not been disclosed, which means Dreamforce this week becomes the forum where that monetization question gets answered publicly for the first time.
What Could Go Wrong
The risks are real. Gartner reports that over 40% of agentic AI projects will be canceled by the end of 2027 due to escalating costs, unclear business value, or inadequate risk controls. Salesforce is betting that enterprise inertia and governance anxiety push customers toward a single trusted platform, but Microsoft through Copilot and ServiceNow are making identical arguments to the same IT decision-makers.
The stock reflects that uncertainty honestly. Salesforce shares trade around $248, with a market capitalization near $203 billion. Over the twelve months through September 2, 2026, CRM’s total return was about 2.5%.
A disciplined long-term investor would ask whether the Trusted Enterprise AI Harness converts into a recurring, high-margin revenue stream, or whether it remains a retention tool that supports existing seat revenue without adding a material new line. The answer is not visible yet. What is visible is that Salesforce understands the question, and this week in San Francisco is its best chance to answer it on its own terms.