The story the industry keeps telling itself is that per-seat software pricing is dead, killed by AI agents that do not need a seat. I do not believe it, and the biggest AI licensing deals signed this year are the evidence. When enterprises actually commit real budget to agentic AI, they are not asking for metered, pay-per-outcome pricing. They are asking for the opposite: a flat number they can put in a budget line and defend to a board, which is exactly what Salesforce is now selling at scale.
The Case for Consumption Pricing, Stated Fairly
The argument for metered AI pricing is a good one and I want to state it before I argue against it. Zendesk was among the first to price its AI agents on a per-resolution basis: the customer pays only when the agent actually resolves an issue, not for a seat that might sit idle. That is a genuinely attractive pitch to a CFO. It ties cost directly to value delivered, and it protects a buyer from paying for AI capacity nobody uses. If software has spent two decades overcharging for unused seats, outcome pricing looks like the correction.
Salesforce itself offers a version of this. Its Agentforce pricing page lists Flex Credits at 500 dollars per 100,000 credits, with individual actions metered by complexity, and a straight 2 dollar per-conversation rate for customer-facing agents. On paper, this is the fair, pay-for-value model the market says it wants.
Advertisement
But the Biggest Deals Are Not Buying It That Way
Look at where Salesforce’s actual enterprise commitments have landed instead. In July, the U.S. Department of Veterans Affairs signed a 1.6 billion dollar, three-year Agentic Enterprise License Agreement, a flat structure that bundles Agentforce, Data 360, MuleSoft and Slack into one committed number rather than a metered bill. Months earlier, the Adecco Group signed a multi-year unlimited global agreement covering Agentforce 360 across its three business units and more than 60 countries, explicitly framed around predictability rather than usage-based billing.
“By moving beyond experimentation to a full-scale agentic enterprise, the Adecco Group is proving that autonomous agents can deliver the determinism and predictability needed to power a global business,” said Madhav Thattai, EVP and GM of Agentforce at Salesforce, on the Adecco agreement. Notice the word he reached for. Not efficiency, not value alignment. Determinism. That is procurement language, and it is the opposite of what a metered, outcome-based bill provides.
Salesforce’s Own Product Admits the Problem
Salesforce did not stop at offering a flat-fee option alongside consumption pricing. It built a Digital Wallet into Agentforce specifically to give buyers near real-time visibility into consumption and proactive threshold alerts before they blow through a budget. A vendor does not build a spend-alarm feature for a pricing model that customers find comfortable. It builds one because unpredictable, usage-metered AI bills are a genuine source of budget anxiety, and Salesforce knows its own consumption pricing creates exactly that anxiety in the accounts that use it.
What It Means for the Sales Leader
If you sell, or buy, agentic AI for a revenue team, do not assume the market is migrating toward pure consumption pricing just because it is the fairer idea in principle. The RevOps buyers now scoring CRM platforms on AI depth are, in my experience of how enterprise procurement actually behaves, going to keep asking the same question they ask about every other line item: what is this going to cost me next year, exactly, in a number I can defend upward. A pricing model that cannot answer that question with a straight number is going to keep losing the largest, highest-stakes deals to one that can, however elegant its per-outcome logic looks in a pitch deck. The same trust gap that is forcing delivery partners to prove their AI claims before enterprises will sign is forcing pricing models through the identical filter: buyers do not just want to trust that the AI works, they want to know in advance exactly what it will cost them if it does.
The Bet I Am Making
Vendors still pitching pure consumption or per-outcome pricing as the inevitable future of enterprise AI should look at where their own largest customers actually signed. If the next twelve months bring more billion-dollar AELA-style commitments and fewer marquee consumption deals, that is not a temporary detour on the way to metered pricing. It is the market telling every AI vendor what its biggest customers have already decided: predictability beats precision, every time real budget is on the line.
Source: PR Newswire
