How LLMs Are Destroying 5 Vertical SaaS Moats While 5 Others Survive
A viral analysis by entrepreneur Nicolas Bustamante has laid bare one of the most uncomfortable truths in tech: large language models are systematically dismantling the moats that made vertical SaaS companies worth billions. His framework, which has been viewed over 2 million times, identifies exactly which competitive advantages survive the AI revolution — and which ones don't.
The $1 Trillion Software Stock Selloff Makes Structural Sense
Over the past year, roughly $1 trillion has been wiped from software stock valuations. Many analysts dismissed this as market overreaction, but Bustamante — who built Doctrine, a legal-tech SaaS company, and now runs Fintool in finance — argues the selloff is "structurally justified, but temporally exaggerated."
The core problem: where a vertical SaaS category once had 3 competitors, LLMs could enable 300. The barrier to building software has collapsed, and the question every founder and investor now faces is which moats actually hold.
The 5 Moats That LLMs Are Destroying
1. Learned User Interfaces
Vertical SaaS products spent years training users on complex, specialized interfaces. But when AI can understand natural language, users no longer need to learn a product's specific UI. A text prompt replaces a 50-button toolbar. This moat has been completely neutralized.
2. Custom Workflows and Automation
Building custom workflows used to require deep industry knowledge coded into software. Now, LLMs can generate, modify, and optimize workflows on the fly. What took months of development can be replicated in hours using AI agents that understand the domain.
3. Public Data Access and Aggregation
Many vertical SaaS companies built their value by aggregating publicly available data into usable formats. LLMs can now scrape, structure, and analyze public data at scale, eliminating the middleman entirely.
4. Talent Scarcity
Specialized industries faced severe talent shortages — finding developers who understand both code and, say, maritime shipping regulations was nearly impossible. LLMs bridge this gap by encoding domain expertise, making it vastly easier for generalist developers to build industry-specific tools.
5. Feature Bundling
The classic SaaS strategy of bundling many features into one platform loses power when AI can stitch together best-of-breed solutions seamlessly. Users no longer need a monolithic platform when an AI agent can orchestrate multiple specialized tools.
The 5 Moats That Still Hold
1. Proprietary Data
If your SaaS product generates or captures unique, proprietary data that doesn't exist anywhere else, LLMs cannot replicate it. This is the strongest surviving moat — data that customers create within your platform becomes a genuine competitive advantage.
2. Regulatory Lock-In
Industries like healthcare, finance, and legal operate under strict compliance frameworks. Getting certified, maintaining audit trails, and meeting regulatory requirements creates switching costs that LLMs alone cannot overcome.
3. Network Effects
When a vertical SaaS product becomes the platform where an entire industry transacts — like a marketplace connecting buyers and sellers — each new user makes the platform more valuable. LLMs cannot easily replicate an established network.
4. Transaction Embedding
Companies that embed themselves into actual financial transactions (processing payments, managing billing, handling insurance claims) create dependencies that are extremely difficult to displace, regardless of how capable AI becomes.
5. System of Record
Being the authoritative source of truth for critical business data — the system of record — creates massive switching costs. Migrating years of structured, validated data from one system to another is painful enough that most companies simply won't do it.
The Bottom Line
Bustamante's framework offers a practical lens for evaluating any vertical SaaS company's durability in the AI era. If your moat relies on things AI can replicate — interfaces, workflows, public data — you're in trouble. If it relies on things AI cannot easily replicate — proprietary data, regulatory compliance, network effects, embedded transactions, or system-of-record status — you have a fighting chance.
The implication for investors: not all software companies deserve the same discount. And for founders: the time to honestly assess which moats protect your business is now, not after the next wave of AI-native competitors arrives.