From SaaS Seats to Outcome-Based Revenue
Andreessen Horowitz (a16z) is arguing that the economics of artificial intelligence are shifting away from the per-seat subscription model that defined enterprise software for two decades. In a widely circulated thesis, the firm contends that long-horizon AI agents — systems capable of autonomously executing multi-step workflows over hours or days — will push value capture from cost savings toward revenue sharing on results.
The logic is straightforward: if an AI agent can independently shepherd a high-value customer journey to completion, charging a flat software fee leaves enormous value on the table. Instead, the agent’s provider should take a percentage of the outcome it produces.
Where the Model Bites First
The thesis points to insurance, mortgages, and law firms as the initial battlegrounds. These are sectors where a single closed deal can be worth thousands to millions of dollars, where workflows are document-heavy and repetitive, and where labor costs are concentrated in junior staff who spend their days chasing paperwork rather than closing transactions.
- Insurance: Agents that handle claims triage, underwriting prep, and policy binding could be compensated on policies bound or claims resolved.
- Mortgages: Loan origination agents could take basis points on funded loans rather than a monthly license fee.
- Legal: Contract review and diligence agents could bill a share of deal value or hourly savings realized.
a16z frames these as the domains most likely to spawn the next cohort of AI unicorns, because the value of a completed outcome dwarfs the cost of the software that produced it.
Why This Matters for Crypto Rails
Outcome-based compensation creates a measurement and settlement problem. To pay an agent a share of a closed deal, you need verifiable proof that the deal closed, tamper-resistant attribution of which agent contributed, and a payment rail that can split proceeds automatically. That is precisely the design space where on-chain settlement, smart contracts, and verifiable computation have been pitched — and where decentralized compute and agent frameworks increasingly compete.
If a16z is right, the winners will not be the firms with the best model benchmarks alone, but those that can prove outcomes and get paid for them. That favors infrastructure that makes results auditable and payments programmable.
The Forward View
The shift will not happen overnight. Enterprises are culturally wired to buy software, not to hand over a slice of revenue, and legal departments will resist outcome-linked contracts. But the direction of travel is clear: as agents move from assisting humans to replacing entire workflows, pricing will follow the value, not the seat count. The firms that figure out outcome attribution first will define the next software business model — and crypto’s settlement layer may be the quiet beneficiary.
