The question for vertical SaaS used to be: how do I make a better tool for this professional? Julien Bek, writing for Sequoia Capital, argues the question has changed:

If you sell the tool, you’re in a race against the model. But if you sell the work, every improvement in the model makes your service faster, cheaper, and harder to compete with. A company might spend $10K a year for QuickBooks and $120K on an accountant to close the books. The next legendary company will just close the books.

Bek draws a clean line between intelligence work (rule-based execution AI can already handle) and judgment work (experience, taste, strategic calls):

Writing code is mostly intelligence. Knowing what to build next is judgement. […] Deciding which feature to build next, whether to take on tech debt, when to ship before it’s ready.

That split tells product builders where to start: outsourced, intelligence-heavy tasks where a budget line already exists and the buyer is already purchasing an outcome. Replacing an outsourcing contract is a vendor swap. Replacing headcount is a reorg. Start with the swap.

But the part that should reshape how designers think about product strategy is the convergence thesis:

Today’s judgement will become tomorrow’s intelligence. As AI systems accumulate proprietary data about what good judgement looks like in their domain, the frontier will shift. Copilots and autopilots will converge.

This is data recipes given a business model. The moat for the next generation of vertical products won’t be the interface or even the model underneath it. It’ll be the compounding dataset of domain-specific decisions—what “good” looks like in insurance brokerage or medical coding or contract law. Every task the autopilot completes teaches it something the copilot never learns, because the copilot hands that knowledge back to the human.

Bek maps this across a dozen verticals with TAM estimates. Worth reading the full piece if you’re thinking about how to build the next generation of AI tools.