Why proprietary data won't save software from AI

The enterprise software playbook—high margins from switching costs and code defensibility—is collapsing as AI makes building feature-parity products trivial and compute costs scale with usage, not seat licenses. Vendors are pivoting to "proprietary data" as their last moat, but data advantages erode fastest in competitive markets where the best players accumulate users quickly. Most enterprise data isn't actually that defensible. It's often messy, domain-specific, and replicable through alternative sources or synthetic generation. The winners will be those with either network effects that compound faster than data can be replicated, or naked cost advantages in compute—not those betting on information asymmetries that AI is designed to dissolve.