David George argues the “comfortable middle” for software companies is over. Only two credible paths remain. Path 1: accelerate growth by 10+ points through genuinely AI-native products (not chatbots bolted onto old SKUs) in 12-18 months, with 50% of R&D in 4-person pods, token/per-use pricing, and new business models where “if an agent can’t consume and pay for your product autonomously, you probably aren’t there yet.” Path 2: rebuild for 40%+ true operating margins (including SBC) through radical cost discipline — flatten management, standardize implementation, kill committees, raise price where you own workflow/switching cost, let long-tail customers churn. Broadcom under Hock Tan is the existence proof.
The binary framing sharpens what Sell the work, not the tool — model improvements compound for services, against software implies: the “work” path is Path 1 (tokens, consumption, outcomes), while companies that stay on seat-based pricing must be ruthlessly efficient (Path 2) or die in the middle. The moat erosion George identifies — data alone not enough, integrations easier to reproduce, workflow/UI advantages collapsing when agents move across systems — maps directly to the 5 falling moats in LLMs selectively destroy vertical software moats — 5 fall, 5 hold, but adds the competitive consequence: competitors will attack each other’s core modules, not just edges, creating core price pressure. This is Technology helps moat businesses but kills commodity businesses applied to the SaaS middle: companies without durable moats see all AI productivity gains flow to customers as lower prices, forcing the margin path. The growth path’s token-based model is the supply-side complement to B2B becomes B2A — agents become the buyer — when agents are the buyer, per-seat pricing makes no sense because the first AI savings customers realize is labor efficiency (fewer seats).