Jason Lemkin: AI Isn't Killing SaaS. Neglect Is — and Marketo Is His Exhibit A

Jason Lemkin’s argument is specific enough to check: legacy B2B SaaS vendors aren’t losing to AI, they’re self-destructing through neglect, and Adobe Marketo is his case study. A company paying roughly $60,000 a year for the platform experienced 1.5 days of downtime, a failed newsletter send to more than 450,000 subscribers, a proposed 20% price increase, and a broken unsubscribe link left unresolved for over two weeks. Lemkin names four patterns behind it: vendors stopped modernizing for agent-operable APIs, raised prices on declining products, let private-equity owners manage to debt covenants instead of product roadmaps, and blamed the market for stalled growth instead of their own execution. The valuation data backs the stakes — companies growing 20%+ trade around 12x revenue, while those growing under 10% trade near 3x, an ~80% discount for the same underlying business model.

Lemkin made a narrower version of the same argument a week earlier, aimed specifically at point solutions: the old rule that buyers purchase rather than build 90% of the time still holds for complex, integrated platforms with real data moats, but it no longer protects simple, single-purpose tools. His proof there was personal — replacing a $4,000/year newsletter vendor that had “shipped nothing new material in 5+ years” with a custom AI-built alternative in about two hours, saving his Chief AI Officer over 90 minutes a week. He calls the resulting pattern “stealth churn”: usage metrics stay flat right up until a customer quietly finishes building the in-house replacement and cancels at renewal with almost no warning.

Read together, the two pieces are one thesis stated at two altitudes: Marketo is what stealth churn looks like once it’s already happened at enterprise scale, and the $4,000 newsletter tool is what it looks like starting today, one AI-literate customer at a time. For any SaaS vendor selling into these accounts, the defensible position Lemkin’s data actually supports isn’t “AI-proof” — it’s “still shipping.”