Deloitte has put a name to a problem most SaaS product teams have already hit: "tollgating" — the practice of charging for the automated, agent-driven actions that increasingly replace human users in software, and the pricing paradox it creates for vendors.

The piece, published under the title "The pricing paradox of agentic SaaS: What to do about tollgating?", addresses a structural threat to the industry's dominant revenue model. Per-seat pricing assumes humans log in and click things. Agentic AI inverts that assumption: when a bot negotiates the API, triggers workflows, and completes tasks on a user's behalf, seat counts decouple from value delivered.

Why does tollgating create a paradox?

The tension Deloitte identifies runs in two directions at once. Charge nothing for agent-driven usage and you cannibalize your own per-seat revenue as automation shrinks headcounts and login frequency. Charge too aggressively — effectively a toll on every agent action — and you penalize the very automation your product is supposed to accelerate, pushing customers toward competitors or in-house builds.

The paradox is that the more successful your agentic product is, the faster it erodes the pricing base it sits on, unless the model itself changes.

What options do product and pricing teams have?

Deloitte frames the question for leaders deciding how to monetize agentic value without taxing adoption out of existence. The practical menu product teams face includes:

• Usage-based tolls on agent actions, which align revenue with delivered value but introduce unpredictability in customer bills and can punish heavy automation.
• Outcome-based pricing, where customers pay for completed results rather than activity — cleanest in theory, hardest to meter and attribute in practice.
• Hybrid models that keep a platform fee while tolling only high-value agent operations.

Each carries a documented failure mode. Usage tolls create bill shock and procurement friction. Outcome pricing requires measurement both sides will accept, which few vendors can guarantee today. Hybrids risk complexity that confuses buyers more than it reassures them.

Where does this leave product teams?

For PMs, the working implication is that pricing cannot stay a post-launch, revenue-team concern in agentic products. Instrumentation for agent actions — what counts as a billable event, who meters it, how customers audit it — is now a product-design decision, not a finance afterthought.

The conditions under which tollgating works are narrow: customers must perceive the agent action as clearly incremental value, the toll must be predictable enough to budget, and the vendor must avoid double-charging where a human and an agent would otherwise pay twice for the same job. Where any of those conditions fail, tollgating reads as a penalty, and customers route around it.

The full Deloitte analysis is worth reading in the original, as the firm lays out its recommendations in detail beyond what a summary can carry.

As agents take over a growing share of software interactions, expect pricing architecture — not features — to become the defining competitive question for SaaS product teams over the next several release cycles.