Here’s the lie that’s funding half the AI startups in Silicon Valley right now: “We automate 75% of the work.”
Investors hear that and see margin. Scalability. Exit multiples. They see a software company with a cost structure that bends toward zero.
What they don’t see is the 25% that can’t be automated. And that 25% is the only thing that will determine whether the company survives its next funding round.
I watched this play out with a founder who built an AI-powered legal document review platform. The tech was brilliant. It digested thousands of contracts in minutes, flagged clauses, suggested edits. The automation rate was 80%.
But the remaining 20% — the final sign-off, the judgment call on a ambiguous clause, the conversation with a nervous client — required a senior partner. That partner’s hourly rate doubled. The 20% became the most expensive part of the business.
Automation doesn’t eliminate the human; it concentrates the value of the human into a smaller, more expensive, and more strategic function.
This is the part nobody’s underwriting. The AI-native pitch deck shows a beautiful funnel of automation. The cap table doesn’t account for the fact that the last 25% of work is not a temporary inefficiency — it’s a permanent, valuable, and increasingly scarce asset.
One reader called this analysis “corporate speak devoid of substance.” I get it. The phrase “the 25% nobody’s underwriting” sounds like a consulting deck from 2018. But the substance is real. It’s the difference between a startup that scales and a startup that implodes when the AI gets good enough to commoditize the first 75%.
Here’s the uncomfortable truth: As AI improves, the remaining human work becomes more valuable, not less. The easier the automation, the rarer the judgment. The cheaper the algorithm, the more expensive the trust.
We’ve been taught that AI replaces people. The reality is messier: AI replaces the easy 75% and makes the remaining 25% harder, more valuable, and harder to hire for.
That means the moat of an AI-native company isn’t the technology. It’s the team’s ability to exercise judgment, build trust, and take accountability for the edge cases. That’s a human capital problem, not a software problem.
If you’re investing in AI startups, start asking one question: “Who owns the 25%? And how do you make sure they don’t leave?”
If you’re building one, stop marketing “75% automation.” Start showing how you retain, train, and amplify the human layer that makes the rest possible.
Because the next unicorn won’t be built by AI. It will be built by the humans who master the messy 25%.
FAQ
Q: Isn't this just a fancy way of saying that humans are still needed? That's obvious.
A: It's not about stating the obvious. The nuance is that the human role becomes more scarce and more valuable as AI scales. Most investors assume the human cost will shrink or become a commodity. The opposite is happening: the 25% becomes a bottleneck that determines scalability and defensibility.
Q: What's the practical implication for founders and investors?
A: For founders: stop marketing your AI startup as fully automated. Highlight the human judgment layer and how you retain it. For investors: start underwriting the 25% — the quality of the team's judgment, trust, and accountability. That's your real moat, not the algorithm.
Q: What's the contrarian take? Will the 25% eventually be automated too?
A: The contrarian take is that the 25% will not be automated away. Instead, it will become the core of the business. The AI part becomes a commodity. The human part becomes the differentiator. Betting that the 25% shrinks is betting against the increasing value of contextual judgment in a world of abundant computation.