The next trillion-dollar company will sell outcomes, not software. Sequoia's Julien Bek argues that founders building AI tools are losing a race against the model providers themselves. The winning move: stop selling the tool, start selling the work. A company pays $10,000 a year for QuickBooks and $120,000 for an accountant to close the books. The company that just closes the books captures both budgets.
The framework that makes this readable is the intelligence-versus-judgement split. Translating a spec into code is intelligence. Deciding what to build next is judgement. Software engineering crossed the AI threshold first, which is why it accounts for over half of all AI tool usage across every profession, while every other category sits in single digits. Bek maps this across a two-axis grid: intelligence-to-judgement ratio on one axis, outsourced-to-insourced ratio on the other. The companies that win start where tasks are already outsourced and intelligence-heavy, because the budget line exists, the buyer already purchases an outcome, and the substitution is a vendor swap, not a reorg. Crosby started with NDAs for exactly this reason.
The piece is worth reading in full for the opportunity map it builds across every major services vertical, each plotted with estimated labor TAM. The copilot-versus-autopilot distinction, and specifically how copilots eventually converge into autopilots as proprietary data accumulates, is the strategic argument that will matter most to anyone building or funding in this space right now.
[READ ORIGINAL →]