A Simple Framework for Identifying AI-Resistant SaaS MoatsAI has fundamentally changed the economics of software. Code is cheaper, faster to produce, and increasingly easy to customise in-house. As a result, many of the traditional moats that once protected SaaS businesses are weakening. Feature sets can be replicated, interfaces can be rebuilt, and even “AI-powered” differentiation is quickly commoditised.Unlike the big headlines suggest, this does not mean all SaaS businesses are equally vulnerable. It means the nature of defensibility has shifted.Here’s the frameworkWhen assessing whether a SaaS business can withstand AI disruption, I find it useful to step back from features and models and ask four more fundamental questions. If a business fails most of these, it is likely competing on borrowed time.The Vault – Do they own the “system of record”?The strongest SaaS businesses are not just interfaces or productivity layers. They are the authoritative source of truth for a critical dataset. When accuracy, history, and auditability matter, customers are far less willing to replace the system, even if a cheaper or more flexible alternative exists.The Shield – Do they provide legal or compliance cover?In many industries, software is not just a tool but a risk transfer mechanism. Customers are paying for accountability, regulatory alignment, and someone to blame when things go wrong. AI can generate answers, but it cannot accept liability. Where this matters, defensibility increases significantly.The Result – Do they charge for outcomes rather than seats?Seat-based pricing assumes software itself is scarce. AI undermines that assumption. Businesses that price based on delivered outcomes, savings, or risk reduction align themselves with customer value rather than usage volume, making them harder to displace by cheaper tools.The Network – Does learning compound across customers?AI is most defensible when insights from one customer improve the product for all others. If each customer’s data only benefits them in isolation, the software is easier to replicate or rebuild internally. Shared learning creates a flywheel that individual in-house solutions struggle to match.Taken together, these four elements act as a quick stress test. AI may make software easier to build, but it does not remove the need for trusted records, accountability, aligned incentives, and collective learning. I believe SaaS businesses that possess several of these traits should face slower, more constrained disruption .With this framework in mind, the following article examines a real-world case where these dynamics can be observed in practice.