SaaS Often Lacks Durable Moats
Jeff Tang, CEO of Athens Research, on Pinecone and the AI stack
The main takeaway is that most SaaS markets stay open because the product itself is rarely hard enough to copy to lock everyone else out. In practice, teams often choose tools that are good enough, easy to adopt, and already have momentum with developers or business users. That leaves room for several large winners in the same category, while durable advantage comes from distribution, brand, embedded workflow, and proprietary data rather than feature novelty alone.
-
Jeff Tang’s point sits in a broader pattern across software categories like project management and observability, where multiple large companies coexist because buyers are not choosing one forever, they are choosing what fits a team’s workflow right now.
-
In AI software, this gets even more extreme because core model capability keeps improving underneath the app. If a startup is mainly a thin layer on top of a model, a better base model can erase much of its product edge in one release cycle.
-
The more defensible layer is what sits around the model, the customer relationships, the integrations into existing systems, the tuning loop from real usage, and the operational workflow where humans and software train the system together over time.
This points toward an AI market with many fast followers and fewer permanent product leads. The companies that compound from here will be the ones that turn distribution into default adoption, gather proprietary workflow data, and become part of how work gets done every day, not just a clever wrapper around the latest model.