Our Take
De Silva is right that per-seat economics collapse when agents replace users, but he undersells the execution risk: vertical specialists still need to prove they can embed themselves deeper than horizontal platforms can copy.
Why it matters
SaaS investors are watching a $300 billion funding wipeout (per Crunchbase reporting cited in the source) signal that horizontal software valuations have peaked. The shift to vertical, AI-native platforms with services baked in will reshape what gets funded and how software vendors price their work.
Do this week
Evaluate your SaaS contract: audit whether your vendor is pricing per-seat or per-outcome, and model what happens to their unit economics if your agent use case cuts seats by 30–50%.
The Per-Seat Model Breaks Down
Richard de Silva, founder of Lateral Investment Management, argues that the SaaS business model that returned billions to Salesforce and Workday investors has hit a structural limit. The trigger is headless AI: when agents perform work instead of humans, companies no longer need a license per employee. A sales team that once required 100 CRM seats may soon operate with 50.
This upends the recurring revenue math that made SaaS predictable and scalable. De Silva identifies three emerging pricing models: per-unit-of-work (a legal AI charges per contract drafted), per-outcome (compliance software takes a cut of recovered overages), or outcome-based (chargeback recovery fees tied to actual recovered value). None of these fit the per-seat subscription envelope.
Generic horizontal platforms are most exposed. De Silva names form builders, project management tools, SMB CRMs, and social schedulers as categories "compressing fast." If an AI agent can autonomously handle the entire workflow, the software becomes redundant.
Defensive Positions Belong to Vertical Specialists
De Silva argues that the winners will possess what he calls the three Ds: Distribution (recurring customer bases), Domain expertise (specialized knowledge of regulated or complex industries), and Data (proprietary, customer-held information inaccessible to frontier AI models).
The moat is not contractual lock-in but economic friction. A legal AI embedded in a law firm's contract repository, an insurance underwriting model trained on years of underwriting decisions, or a bank's loan performance data cannot be exported and replicated elsewhere. Switching vendors means rebuilding domain logic, not migrating a spreadsheet.
De Silva emphasizes that this model merges software and services. Solutions built for healthcare, financial services, construction, cybersecurity, and legal work require human judgment at critical decision points (what the industry calls Human-in-the-Loop). The vendor relationship becomes operational and embedded, not transactional. Every customer engagement generates proprietary data and deepens the competitive moat.
What This Means for Practitioners
Investors should expect the enterprise software market to bifurcate: legacy horizontal SaaS competing on cost and ease, while vertical AI-native platforms compete for labor budgets, compliance budgets, and risk budgets. McKinsey projects a $6 trillion annual productivity opportunity from AI transformation, per De Silva's analysis. Vertical platforms are positioned to capture a slice far larger than traditional per-seat software ever accessed.
For buyers, the implication is that contract renegotiation is coming. Vendors who don't move past per-seat pricing face margin compression. Vendors who embed services as a "compounding asset" rather than a cost center will hold pricing power. The companies that collapse the boundary between software and services entirely will be worth considerably more than their SaaS predecessors—but will require deeper, longer partnerships to justify that premium.