a16z says long-running AI agents could shift pricing from labor savings to outcome-based revenue share

Andreessen Horowitz partner David Haber argues that AI commercialization is moving past narrow automation and seat-based pricing toward a model tied to business outcomes. In his essay, Haber says the next wave of AI companies may not make their money mainly by replacing support reps or selling software seats, but by owning larger portions of high-value customer journeys and getting paid on measurable results. The piece focuses on what a16z calls "long-running agents" — systems that can persist across dozens of steps, multiple communication channels, and timeframes that stretch for weeks or months. Haber points to examples such as plaintiff law firms, insurance claims, mortgage origination, and enterprise sales, where value is created only after a complex sequence of handoffs. In those workflows, he says, a persistent AI agent can reduce drop-off, maintain context, review evidence in depth, and keep a full record across the entire process. Haber also links that technical shift to a commercial one. If an AI system can be tied to revenue growth rather than just cost reduction, vendors may be able to charge by work product or outcomes. The essay uses Red Ventures as a historical case study and lays out three go-to-market paths for long-running agents: selling software platforms, selling services, or launching full-stack AI-native operators.