LLMs selectively destroy vertical software moats — 5 fall, 5 hold

Not all moats are created equal in the LLM era. Bustamante identifies five vertical software defenses that LLMs dismantle: learned interfaces (“LLMs collapse all proprietary interfaces into one Chat”), custom workflows where “years of engineering versus one week of writing” captures the shift to Markdown skill files may replace expensive fine-tuning, public data access that frontier models trivialize, talent scarcity inverted when domain experts encode methodology directly, and bundling broken by agents that orchestrate across providers.

Five moats hold firm: proprietary data that’s irreplaceable, regulatory lock-in, network effects, transaction embedding, and system-of-record status. When moats collapse, the strategic response shifts from selling tools to selling outcomes — Sell the work, not the tool — the services budget dwarfs the software budget, and every model improvement helps whoever sells the outcome argues that model improvements compound for services providers while eroding tool-sellers. This framework refines The UI moat collapses — API quality becomes the purchasing criterion — the UI moat is just one of five collapsing defenses. The resilient moats share a pattern: they’re rooted in data gravity or institutional trust, not in software complexity. This connects to When non-deterministic AI feeds a deterministic system, the deterministic layer governs — software that owns coordination and accountability keeps its place — the SaaS layer that survives is precisely the one anchored in these durable moats. One layer further up the stack is When products and models converge, the moat is the institution — how a company attracts people, distributes authority and compounds judgment — when the product/data moats either fall or hold, the question of what kind of institution surrounds either case becomes the next-tier moat AI cannot copy.