Deloitte has published a Technology Spotlight titled Accounting for Outcome-Based Pricing in an Agentic AI Software Product, signaling that one of the Big Four's advisory practices is now publishing formal guidance on a pricing model that AI-vendor product teams have been adopting experimentally for more than a year.

The piece lands as software vendors selling agent-based systems — autonomous AI products that execute multi-step tasks on behalf of users — confront a recurring finance question: when revenue depends on whether the agent delivers a specific business outcome rather than whether a human user logs in, how should the contract be booked?

What does Deloitte's framing change for PMs?

The spotlight's title alone defines the scope. Deloitte is treating outcome-based pricing not as a marketing variation on usage-based pricing, but as an accounting category in its own right, applied to a product class — agentic AI — that only recently reached commercial form. For product managers building or selling agent platforms, the implication is that finance, audit, and revenue recognition teams will increasingly ask for the same evidence they receive from any results-dependent software deal: contract terms, performance measurement methodology, and dispute resolution mechanics.

Product managers who treat outcome-based contracts as a sales-team preference will find that accounting treatment, not the AI capability itself, becomes the binding constraint on deal structure. Three terms recur in any ASC 606- or IFRS 15-aligned framing of such contracts:

• Outcome definition — the specific, measurable result the vendor commits to (a resolved support case, a closed-won deal, a passed compliance check).
• Performance obligation — whether the vendor's promise is to produce the outcome or merely to provide the agent that may produce it.
• Variable consideration estimate — the constraint on revenue recognition when a portion of consideration depends on a future event the vendor cannot yet confirm.

Where does the framework break?

Outcome-based pricing fails when the underlying outcome cannot be measured cheaply, credibly, or by a third party. A vendor that promises to "increase customer retention" without a clean baseline or counterfactual measurement will find auditors treating the entire contract as one variable-consideration estimate, often with a material constraint applied. The same applies when the agent's contribution to the outcome cannot be isolated from customer-side behavior, or when the agent acts in environments outside the vendor's telemetry.

The accounting friction is highest in three cases product teams should expect to hit within the next two quarters:

1. Contracts priced per task completed where completion criteria are defined by the customer.
2. Bundles that mix a fixed subscription with an outcome bonus, blurring the performance obligation boundary.
3. Multi-agent workflows where attribution of the final outcome to a specific vendor component is non-deterministic.

What product managers should do on Monday

Treat the pricing model as a product decision, not a contract-template decision. Three operational moves follow:

1. Decide the measurement source before the first deal — who records it, who audits it, and what the dispute path is.
2. Align the agent's success metrics with the contract's outcome metric so engineering telemetry and revenue recognition share one definition of "done."
3. Pressure-test variable consideration estimates with finance early; under ASC 606's constraint principle, revenue can be deferred even after delivery if the estimate is unreliable.

The forward trajectory is clear. As agentic products move from pilots to line-item revenue, the conversation between product, sales, and finance shifts from what can the agent do to how do we prove, in auditable terms, that it did it.