How does AI affect software company moats?
Weakening Moats: Human memory, UI learning curves, public information processing, and code volume.
Key Points
- AI agents reduce the cost of writing software toward zero and shift the primary user base from humans to agents, fundamentally altering the nature of competitive moats.
- Moats based on human cognitive friction—such as UI learning curves, brand recall, and the effort of vendor discovery—are eroding. SaaS companies relying on these are at high risk.
- Network effects are weakening because agents can efficiently manage listings and searches across dozens of networks simultaneously, reducing the ‘winner-take-all’ advantage of a single platform.
- Switching costs are decreasing as business context moves from proprietary software silos into owner-controlled markdown files and databases, allowing agents to migrate between vendors in hours.
- Distribution moats are weakening because AI agents can perform exhaustive research across hundreds of vendors, bypassing the ‘top-of-mind’ advantage of dominant brands.
- Branding remains a moat only for ‘hard-to-verify’ work (e.g., enterprise security like CrowdStrike) where the cost of failure is too high to risk on an unproven vendor.
- Economies of scale are shifting: software-only scale is eroding, but capital-intensive infrastructure (data centers, GPUs, physical logistics) remains a durable advantage.
- Proprietary data moats are strengthening, specifically through ‘agentic loops’ where companies use real-time customer interaction data to auto-improve their codebases.
- [AI Synthesis] The shift suggests a transition from ‘Software as a Service’ to ‘Outcome as a Service,’ where the value migrates from the tool (the SaaS) to the entity that owns the proprietary data and the trust layer.