20VC: The 8 Moats of Enduring Software Companies: How to Analyse for Durability and Defensibility in a World of AI | Why Dropouts are "AI Maxing" the World & Remote Early-Stage Companies are Dying with Gokul Rajaram

What It Covers

Gokul Rajaram, angel investor and Marathon founder, presents an eight-moat framework for evaluating software durability in an AI-disrupted market. Drawing on operator experience at Google, Facebook, Square, and DoorDash, he analyzes which software companies survive commoditization, how vertical SaaS must evolve, and why remote early-stage teams and single-product companies face structural disadvantages.

Key Questions Answered

• •Eight-Moat Scoring Framework: Score any software company across eight moats: data, workflow, regulatory, distribution, ecosystem, network, physical infrastructure, and scale. Assign one point per moat. Companies scoring four or above are structurally secure. Two to three signals weakness. One or below requires urgent moat-building. Pure software companies realistically only qualify for data and workflow moats, making those two the primary evaluation criteria at early stage.
• •Vertical SaaS Survival Requires Full-Stack Ownership: Single-function vertical software tools — voice agents for dentists, chiropractors, auto dealers — are viable but unlikely to exceed modest scale. ServiceTitan reached sub-$10B with 32 products serving field services. To build a $10B+ vertical company, founders must own the entire software stack for their vertical and target both BPO spend and human labor budgets, not just existing software line items.
• •Multi-Product Retention Strategy: Square's North Star metric became median number of products used per merchant, not revenue. More products per customer directly correlated with higher retention. Critically, not every product needs to generate profit — some products serve retention while others serve the profit pool. Confusing these two roles causes teams to optimize for the wrong outcomes. Product two must emerge naturally and adjacently from product one.
• •AI Labor Budget Displacement Sequence: Enterprise AI spend is transitioning from software budgets to human labor budgets in a predictable three-stage sequence. First, companies cut third-party BPO contracts (call centers in India and Philippines), where AI delivers higher quality at roughly 30% lower cost. Second, vacated roles go unfilled. Third, layoffs occur. Goldman Sachs and Barclays each employ over 30,000 people in India — that spend represents the primary near-term AI displacement opportunity.
• •Pricing Model Bifurcation: Software products split into two categories requiring different pricing architectures. Access products — where value comes from using the tool — suit seat-based pricing with tiered functionality. Work products — where value comes from output produced on the user's behalf — require outcome-based pricing tied to contracts processed, calls handled, or tasks completed. Charging per seat for a work product misaligns incentives because the user count is no longer the relevant constraint.
• •IRR Over MOIC for Liquidity Decisions: Early-stage fund managers systematically over-optimize for MOIC while ignoring go-forward IRR. One LP cited a firm delivering 7x MOIC over 20 years — a teens-level IRR that underperforms expectations. At each liquidity event, calculate whether the go-forward IRR on remaining position exceeds fund target returns. If a single position represents 20–40% of fund value, selling a portion is an LP obligation regardless of long-term conviction.