Harvey beats Copilot in Repsol pilot

The author argues that AI application companies must establish robust pricing levers—such as superior workflow execution, exclusive context, and demonstrable value—to protect their margins against improving general-purpose models. Drawing on examples like Harvey and Decagon, the analysis emphasizes that applications must outperform customer-built alternatives and price based on recognizable units of completed work. Founders are advised to measure full delivery costs and share efficiency gains while keeping their overall offers attractive. Open article →