Why AI Won't Kill SaaS: The Bread Paradox

Westenberg uses an unused bread machine as a lens on the argument that cheap AI coding tools spell doom for SaaS. Making bread at home is trivially cheap and fully automated, yet Americans still buy roughly 10 million pre-made loaves a day—just as Romans and medieval townsfolk bought from professional bakers rather than baking themselves. The reason isn’t cost of ingredients but the accumulated mental overhead of planning, maintenance, and inconsistency. People don’t pay for the loaf; they pay to not think about it.

The same logic applies to software. AI can now spin up a custom CRM or dashboard in an afternoon, but building it makes you the baker: you inherit maintenance, edge cases, compliance audits, security gaps, and the 2 a.m. outage call. Westenberg notes AI-generated code carries roughly 1.7x the major defects of human-written code, and institutional knowledge evaporates once the person who built the tool moves teams. What durable SaaS vendors actually sell is convenience, consistency, accountability, and someone else to blame when things break.

The caveat is that not all SaaS is safe. Thin, single-function products—PDF-to-spreadsheet converters, meeting summarizers, automated follow-up emailers—are genuinely exposed, since their whole value is a feature a model can now replicate from a prompt. The companies that survive are the industrial bakeries of software: deep integrations, proprietary data, regulatory certifications, years of domain logic, and partner ecosystems. Your ability to bake a loaf never threatened commercial baking, and your ability to prompt an app won’t threaten mature SaaS.