The Liability Line Nobody Mentioned

Julien Bek's "Services: The New Software" makes the case that the next giant AI companies won't sell tools. They'll sell the work itself. Copilots sell software. Autopilots sell outcomes. The money is in outcomes because for every dollar spent on software, six are spent on services. It's a Sequoia deal memo wearing essay clothes, and it's worth reading for one specific move that most AI strategy writing doesn't make.

That move is the outsourcing-as-wedge test. Bek argues that if a task is already outsourced, you know three things about it: the company accepts external delivery, a budget line exists, and the buyer is already purchasing an outcome. That isn't a theory. It's a procurement filter. It tells you exactly where an AI service company faces a vendor swap instead of a reorg. Most founder advice about market entry is vague. This is a concrete, falsifiable heuristic you can apply to any vertical in ten minutes.

Here's what the piece doesn't earn: the intelligence/judgement spectrum that holds the entire argument together.

Bek draws a clean line. Intelligence is rules-based work: translating a spec into code, filling insurance forms, coding medical bills. Judgement is taste and experience: knowing what to build, advising a client, assessing culture fit. AI handles intelligence now. Judgement comes later. Autopilots start with the intelligence-heavy work and expand from there.

In practice, intelligence and judgement are tangled inside the same task. An NDA is "pure intelligence" until one clause depends on understanding the client's commercial leverage over a counterparty. Tax compliance is "mechanical" until the planning decisions baked into the filing aren't. The spectrum works as a directional heuristic. It doesn't work as the load-bearing structure for a trillion-dollar thesis.

You can see the framework buckle. At the bottom of the opportunity map, management consulting — the single largest market listed, at $300–400B — gets a one-paragraph shrug: "mostly judgement," "best candidates TBD." That isn't a minor footnote. It's the argument hitting its own ceiling and quietly moving on before anyone notices.

The bigger gap is liability. When you sell a tool, the professional owns the output. When you sell the work, you own it. Bek lists insurance brokerage, healthcare billing, and legal work as prime autopilot territory, all heavily regulated, all carrying real consequences when the work is wrong. The shift from "we help your people do the work" to "we do the work" isn't just a go-to-market decision. It's a regime change in who gets sued. The piece treats distribution as the hard problem and responsibility transfer as a detail. In regulated verticals, it's the opposite.

Read it for the outsourcing wedge. Discount the spectrum. And notice what's missing on liability, because that's where the actual difficulty lives for every company on the map.