SaaS Durability in the Age of AI

• LLMs make custom software trivial to generate; the moat must live outside the software layer.

  Workflow logic, form-based data entry, and being a thin UI over a database are no longer defensible — these are the categories AI agents synthesize fastest. Durable SaaS franchises are protected by something that is not code: proprietary data, regulatory and compliance entrenchment, two-sided network effects, hardware and calibration loops, or mission-critical blast radius. The framework deliberately ignores valuation and management quality. The question is which businesses are structurally protected, not which are cheap.

• Five moat categories with different decay curves.

  (1) Proprietary Data — decades of curated datasets AI cannot regenerate without the underlying observations. (2) Regulated Systems of Record — compliance, audit trail, and regulatory relationships create switching costs unrelated to software quality. (3) Two-Sided Networks — ecosystem value lives beyond the UI. (4) Physical-World Integration — hardware, sensors, and calibration loops code alone cannot replicate. (5) Mission-Critical Infrastructure — tiered by severity of failure.

• Mission-Critical Infrastructure is the largest and deepest bucket — sub-tiered by blast radius.

  Existential failure threatens the existence of customer operations; switching is essentially unthinkable. Operational failure severely disrupts daily operations but is recoverable. Engineering-Critical switching costs are measured in years because the tools sit inside multi-year product development cycles. The CIO test applies cleanly: if the board would ask follow-up questions about risk and continuity before approving a vendor change, the name is durable.

• Cybersecurity is the underrated tailwind — AI raises both attack sophistication and attack surface.

  Defenders gain because attackers gain. The Cybersecurity Tailwind sub-theme is cross-cutting under SaaS Durability rather than under Mission-Critical because the driver is demand-side, not moat-side — and the same names also carry distinct edges to PQC Migration (deadline-driven) and Existential Infrastructure (blast-radius driven). Same companies, three distinct economic forces.

• Contrarian take — the biggest winners may be incumbents that reposition as the AI agent orchestration layer.

  Agentforce and ServiceNow's AI platform are early bets that customers will bring any LLM as the interface while incumbents own the data, permissions, audit trail, and integrations beneath it. This is the AI Orchestration Layer Incumbents thesis as a separate theme, but the names live first under SaaS Durability. If the orchestration thesis plays out, these go from Significant to Core strength on the durability framework.

• Hyperscaler-led customer migration off a thin-moat horizontal SaaS vendor.

  First well-documented case of a Fortune 500 customer cancelling a thin-moat horizontal vendor in favor of an internal LLM-built equivalent would be the inflection point. Watch press releases, cloud-provider customer case studies, and earnings-call commentary on churn. Would validate the durability thesis (winners stay) and the disruption thesis (losers get hit) simultaneously.

• AI orchestration layer revenue disclosure from CRM or NOW.

  Both have launched AI platform products; neither has broken out attributable revenue. The first quarter either reports a material attributable line item confirms the contrarian take. Conversely, repeated guidance disappointment would undermine the orchestration thesis.

• Multiple compression in horizontal SaaS opens a relative-value gap.

  If the AI Disruption headwind theme plays out, thin-moat names should de-rate relative to the durable cohort. A wider multiple gap creates the conditions for premium durable names to look expensive vs. peers — which is the point. The framework predicts the dispersion.

• Major SaaS M&A by a Mission-Critical or Proprietary Data incumbent.

  Roper is the canonical compounder buying niche mission-critical software at scale; S&P Global's IHS Markit deal demonstrated the Proprietary Data play. The next $5B+ deal where a durable incumbent acquires a vertical software company at a premium signals the moat-by-acquisition strategy is accelerating.

• Cyber breach attributed to AI-generated attack tooling.

  First major breach explicitly tied to LLM-assisted attack development accelerates the Cybersecurity Tailwind sub-theme materially, driving security spend and board-level demand for defender names across the existential tier.

• Engineering-Critical incumbents ship credible AI copilots without revenue cannibalization.

  These businesses have the deepest moats but the most to lose if a price-disrupting AI workflow emerges in chip design or CAD. Quiet integration of AI features inside existing seat-based pricing — without seat reductions — validates the engineering-critical thesis and could re-rate the cohort higher.

Not exhaustive — see Neo4j for the full edge list.

• Framework, not investment advice. This identifies what is structurally protected from AI substitution, not what is attractively valued. Many durable names trade at premium multiples that already price in some of the moat.
• Same names appear in multiple sub-themes — intentionally additive, not duplicative. VEEV carries three distinct-mechanism edges; the cyber names carry four. Multi-theme exposure inflates apparent overlap in slicing queries.
• The framework is theme-driven, not sector-driven. GICS edges still classify these companies, but theme exposure is a separate analytical layer. Don't infer durability from sector.
• Mechanism is required on every edge. The graph currently has Mechanism populated at the L1 parent but missing on several L2 child edges — pending a backfill pass.
• The AI Orchestration Layer Incumbents thesis lives as a separate L1 theme, not a sub-theme here. Revisit promotion to a Level 2 child only if revenue disclosure validates the orchestration take.
• Engineering-Critical names are the deepest moat but the highest risk if AI changes the workflow primitive. Watch for early signals of seat reductions on AI feature ship.