Token Usage Outcome Pricing Captures AI Growth
Claim
The fastest-growing AI software budget will shift toward tokens, consumption, automations, outcomes, and machine-driven workflows rather than classic human-seat packaging.
Raised by
- 2026-03-23-a16z-two-paths-left-for-software - david-george / a16z argues companies need to be “in the token path” to stand in the fastest-growing part of the budget.
- 2026-03-05-sequoia-services-the-new-software - julien-bek / sequoia-capital extends the pricing logic from tokens/usage toward completed work and services outcomes.
Supporting evidence
- 2026-03-23-a16z-two-paths-left-for-software - says new growth will increasingly sit in tokens, consumption, automations, outcomes, and machine-driven workflows.
- 2026-03-05-sequoia-services-the-new-software - argues that autopilots sell work directly, capture work budgets from day one, and should start where buyers already purchase outcomes.
Counter-evidence
- (unknown - needs source) Usage and outcome pricing can be harder to procure, forecast, and attribute than seats; customers may resist it even if value creation shifts there.
Implications
- Product design should expose measurable units of work: calls analyzed, accounts managed, renewals rescued, campaigns executed, workflows completed, or agent actions taken.
- Outcome pricing may need service-level guarantees and QA, not just token metering, when the product sells completed work.
- Agent-readable APIs, logs, evals, and pricing meters become part of the product surface, not back-office details.
- The team should design pricing experiments before overbuilding product, because the pricing unit may determine the wedge.
Ideas this favors
- agent-native-go-to-market - because the product should be consumable by agents and priced around machine work.
- autopilots-capture-services-budget - because completed work is the strongest version of outcome pricing.
- full-stack-ai-vertical-services - if the company captures the services budget directly instead of selling software into it.
- account-management-vertical - if the unit becomes managed account coverage, meeting prep/follow-up, renewal risk workflows, or portfolio expansion.
- revenue-cycle-brain - if the unit becomes closed-loop GTM actions and measured revenue impact.
Ideas this weakens
- Products whose only unit is “number of human users with access.”
- Internal productivity tools with no measurable output unit.
Confidence
Medium. Strong fit with the team’s existing agent-native thesis, but still needs buyer validation around willingness to buy usage/outcome units in the chosen vertical.
What would change our mind
- Buyers prefer simple seats even for high-volume AI work.
- Outcome attribution proves too noisy for procurement or renewal decisions.
- Token costs fall so much that usage becomes a poor value metric.