The Line We Invest Around: Execution, Judgment, and the Outcome Economy

Venture capital used to have two unbreakable rules: never invest in hardware, and never invest in services. We've made exceptions for the first (physical AI), but the second rule hasn't just bent; it has been entirely rewritten by the Outcome Economy.

The old objection to services businesses was structural: they didn't scale, margins were capped by headcount, and you were really just buying a very good consulting practice with venture-style expectations attached. That objection was never about services being a bad idea. It was about services being a bad match for venture math.

What's changed isn't the nature of services work. It's who, or what, is doing it.

The Outcome Economy: Software vs. Services

Take customer support. A tool that helps a support team answer tickets faster is still a services business wearing a software costume, because you're still paying for headcount, just slightly more efficient headcount.

A company that resolves a guaranteed volume of tickets, and prices accordingly, has actually changed categories. It's no longer selling a better process. It's selling the outcome the process used to require a team to produce.

That is a small-sounding distinction with a massive consequence. Instead of competing for a sliver of an IT budget, you're now competing for a share of headcount, usually the single largest expense on a company's books. If you're looking for where this shift shows up first inside a large enterprise, the answer is simple: look at whatever the company already sends outside its own walls. Anything outsourced or offshored has been pre-qualified. The work is standardized, the company is comfortable buying a result instead of doing it internally, and there is an established budget. Replacing an outsourcing contract with an AI provider isn't asking anyone to accept a new paradigm; you're just asking them to accept a better, faster version of a decision they made years ago.

The Line: Execution vs. Judgment

This is where the real line we invest around comes into focus. AI is genuinely excellent at mechanical, well-specified execution. It is still not something we trust with a judgment call that has no clean right answer, the kind of decision a good operator makes by weighing context nobody wrote down.

Almost every workflow is actually two workflows stitched together: a repetitive, rules-based one, and a smaller judgment-based one sitting on top of it. Every company we're excited about in this space sells the first piece completely and is honest about not touching the second.

Our portfolio perfectly reflects this Execution vs. Judgment divide:

• Signal Labs built a Systems of Attention platform to triage incoming enterprise signals. It doesn't make the strategic call; it simply guarantees that the person who does make the call is looking at the right thing at the right time.

• Collective gives a solo entrepreneur a finished tax function instead of just selling them tax software. It executes the mechanical compliance outcome, leaving the entrepreneur to exercise their actual craft.

• Nas.com acts as a "business in a box" for e-commerce founders. It completely takes over the back-office and marketing execution, allowing the founder to focus purely on creative judgment.

In each case, the company took over an entire chunk of work a person used to have to do themselves, instead of just making that person marginally faster at doing it.

The Unintended Consequence: The Apprenticeship Crisis

One level in from there, the line gets harder to draw. We used to assume expertise-gated professions like law, banking, and software engineering would sit safely on the judgment side, since the work requires years of training no shortcut replaces.

What's actually happening isn't a split between protected and exposed professions. It's the same structural failure showing up everywhere: the routine work juniors were assigned was never separate from their training; it was the training.

Document review taught junior lawyers how contracts actually break. Debugging taught junior engineers how systems actually fail. When that work gets automated, the apprenticeship doesn't get more efficient. It disappears. There was never a training program running alongside the grunt work. The grunt work was the program.

Any individual firm captures the full benefit of skipping junior training today: lower costs and faster output, immediately. But the cost of a future talent shortage is delayed and distributed across the whole industry. A firm that trains junior people today is paying for a benefit competitors can just as easily poach tomorrow. Because of that imbalance, no single firm has the incentive to keep training people just to prevent a problem everyone will eventually share.

It is a real due-diligence question, because a startup that automates pure throughput with no plan for where its own customers' future judgment comes from is building toward a ceiling. Right now there's an obvious, easy pool of throughput work to automate, and that's driving a lot of the early growth in this category. But that pool isn't infinite.

The next tier of product these startups will eventually want to sell is closer to judgment: tools that need an experienced person on the other end to direct them and catch their mistakes. If a startup's own product has already helped eliminate the junior staff who would have grown into those experienced roles, its own customers may not have anyone left capable of buying that next product.

The startup ends up competing for a shrinking pool of judgment-capable buyers, some of that shrinkage caused by its own earlier success.

We ask founders whether they've thought about that, not because we expect them to fix the labor market, but because the ones who have usually have a better second act than "sell more seats of the same automation."

The Takeaway

None of this is a claim that judgment is safe forever, only that it's safe for now, and "for now" is doing real work in that sentence.

The founders we're most drawn to are the ones who've clearly thought about where that line currently sits, rather than those who find it convenient to pretend the line doesn't exist. The companies that respect it and sell concrete outcomes are going to be genuinely massive. The ones that quietly blur it tend to get found out, usually by a customer who assumed a person was checking the work, and found out the hard way that no one was.