Salesforce just changed how AI agents get sold. On June 25, the company launched Agentforce Help Agent with a pricing model that most enterprise vendors have avoided: pay-per-resolution. You only pay when the agent actually resolves a customer issue from start to finish. If it fails, if it escalates, if it half-answers and hands off to a human, you pay nothing.

This is not a minor product update. It is the largest CRM company on earth betting its revenue on whether its AI actually works. And the number backing it up is real: Salesforce reports Agentforce resolved 70% of 4.3 million customer inquiries on its own support portal without human intervention.

That is roughly 3 million issues handled autonomously. And now they are telling enterprises: we will only charge you for the ones we solve.

The pricing model is the strategy

Most AI vendors still sell the way software has always been sold. Per seat. Per user. Per month. You pay for access to the tool whether your team uses it well, uses it badly, or barely uses it at all.

Pay-per-resolution flips that. It transfers performance risk from the buyer to the vendor. If the agent does not perform, Salesforce does not get paid. That is a different kind of relationship than any seat-based license has ever created.

According to Futurum Group’s 1H 2026 Enterprise Software Decision Maker Survey (n=830), 18.7% of enterprises now use outcome-based pricing tied to agreed-upon metrics, approaching parity with per-user/per-month models at 17.1% for the first time. And 52.2% of buyers cite pricing model as a primary purchase decision criterion.

The shift is not theoretical. Buyers are already voting with procurement decisions. Salesforce just made it impossible for its competitors to ignore.

Why this matters more than the technology

Here is what most of the coverage is getting wrong. The interesting part of this announcement is not the AI. Customer service automation has existed for years. Chatbots, intent routing, knowledge base retrieval. None of that is new.

What is new is the business model underneath it.

When a vendor prices on outcomes, it forces a specific kind of operational discipline. The agent has to actually work inside your systems, across your channels, against your real customer problems. Not in a demo. Not in a pilot. In production, handling real volume, resolving real issues.

Salesforce is claiming its Help Agent deploys in minutes with guided setup across voice, web, portal, and messaging. That is a simplicity play. The barrier to entry drops to nearly zero. And the risk drops to literally zero because you do not pay unless it performs.

For a business leader evaluating AI tools, this changes the entire conversation. The question stops being “can we afford to try this?” and becomes “can we afford not to, when there is no downside?”

The competitive pressure is immediate

Salesforce is not the first company to experiment with outcome-based AI pricing. Decagon, Intercom, and Zendesk have all tested versions of resolution-tied pricing in the customer service space. Adobe, Automation Anywhere, ServiceNow, and UiPath have introduced hybrid models blending subscription with outcome-aligned billing.

But none of them are Salesforce. None of them have 150,000+ enterprise customers and the installed base to force an industry standard.

When a company at that scale commits to this model, it legitimizes it for every buyer in the market. And it puts immediate pressure on Microsoft, ServiceNow, and every other platform vendor still selling copilots and agent frameworks on traditional pricing. The question buyers will now ask their existing vendors is simple: if Salesforce will only charge me when it works, why are you charging me regardless?

That is a question most vendors are not ready to answer.

What this actually means for your business

If you run customer service, support, or any high-volume interaction workflow, this is worth paying attention to right now. Not because Salesforce built a better chatbot. Because the economics of AI just shifted in the buyer’s favor for the first time.

Pay-per-resolution means you can deploy an AI agent with zero financial risk. If it resolves 70% of your volume the way Salesforce claims, you are paying for results you would have staffed humans to produce. If it resolves 30%, you pay for 30%. If it resolves nothing, you pay nothing.

That is not a technology decision. That is an operations decision. And it is the kind of decision that separates organizations moving on AI from organizations still evaluating it.

The real signal here is not what Salesforce built. It is what Salesforce is willing to stake its revenue on. When the vendor’s incentive aligns perfectly with the buyer’s outcome, the adoption barrier disappears.

Every enterprise software company watching this launch is running the same numbers right now. Some will match the model. The rest will spend the next two years in sales calls trying to explain why their AI is worth paying for whether it works or not.

Good luck with that pitch.