A flat fee that isn't what it looks like
Salesforce spent most of 2025 trying to figure out how to charge for Agentforce. First it was Flex Credits, a consumption model priced against tokens and actions. Then it added a per-conversation rate for customer-facing bots, at $2 a conversation. Then a $125-per-user add-on for unmetered employee use. None of it stuck as the single answer, and none of it is what's on the table now.
The current offer is the Agentic Enterprise License Agreement, AELA for short. It bundles Agentforce with Data 360, MuleSoft and parts of Slack into one multi-year contract, priced as a fixed fee rather than metered by usage. Sign it, and a customer gets unlimited use of those products for the length of the term, typically two to three years, with no per-conversation charge and no credit meter running in the background.
Salesforce's own pricing update, published on its newsroom in late 2025, frames this as simplification: predictable costs instead of a bill that moves every quarter. Forrester, writing about the shift in its product-pricing blog, described it more precisely: AELA reframes AI agents as economic actors that generate value, not metered features that carry variable cost. That is a generous way of saying Salesforce stopped trying to price the thing it's selling and started pricing the relationship instead.
The quote that explains the strategy
The part of this story that should get more attention than the pricing mechanics is what Salesforce's own leadership has said about it. Chief Revenue Officer Miguel Milano, speaking at a conference covered by The Register in December 2025, said the company is comfortable losing money on individual AELA deals in the near term. His reasoning, as reported: if a customer uses agents so heavily that Salesforce loses money on the contract, that customer is by definition the happiest one in the portfolio, and Salesforce has decades to monetize the relationship it just built.
That is not a hedge. It is a stated plan to underprice a product deliberately, in order to get agentic workflows wired into a customer's operations before the true cost of running them is priced back in. The flat fee is not a gift. It is customer acquisition cost paid in the form of a discount, recovered later once the workflows are load-bearing and switching them out means rebuilding automation, integrations and headcount plans built around Agentforce specifically.
Why "unlimited" rarely means unlimited
Buyers who have negotiated AELA terms report that the word doing the least work in the contract is "unlimited." Licensing analysts who track these deals note that most AELA agreements carry fair-use language, baseline usage assumptions tied to the customer's stated use case at signing, and true-up provisions that trigger if actual usage runs materially past that baseline. None of this is unusual for enterprise software, multi-year enterprise agreements have carried true-up clauses for decades. What's different here is what triggers the true-up: agentic usage volume, which is inherently harder for a buyer to forecast at signing than seat count ever was. A company buying 500 CRM seats can count desks. A company buying "unlimited" agent usage across Data 360 and MuleSoft is estimating a number that depends on how aggressively its own teams adopt automation over a term that runs two to three years past the signature.
The buyer impact
For a CIO or procurement lead evaluating this deal, three things matter more than the headline flat rate.
First, budget certainty is real, but it is certainty about one number, the base fee, not certainty about total cost. The true-up mechanics mean the contract can still produce a surprise bill, just later and larger than it would have under per-conversation pricing, because usage has had two to three years to compound before it's re-priced.
Second, the multi-year term is the actual product being sold. Salesforce's willingness to lose money up front only makes financial sense if the term is long enough, and the workflows deep enough, that switching cost becomes structural before the contract is up for renewal. A buyer signing a three-year AELA should assume the renewal conversation will start from a position where walking away means re-platforming live agentic workflows, not swapping a CRM seat count.
Third, this is a template other vendors will copy, which means the leverage question won't stay confined to one supplier. If flat-fee, loss-leading agentic licensing becomes the standard entry point across the CRM, data platform and integration layers a company runs, procurement teams lose the one lever that has worked against per-seat inflation for the past decade: multi-vendor comparison shopping on price per unit of usage. A flat fee across a bundle of products from one vendor is much harder to benchmark against a competitor's flat fee across a different bundle.
What to ask before signing
Before signing an AELA or any comparable agreement from another vendor, a buyer should get three things in writing rather than in the sales deck: the exact usage baseline the flat fee assumes, the formula and cadence for any true-up, and an explicit exit or downgrade path if agentic adoption underperforms the vendor's own projections. None of those questions are hostile to the vendor. They are the difference between buying predictability and buying a bet that someone else priced in their own favor.
Sources: Salesforce pricing update · Forrester · The Register · RedressCompliance