Flatter, Faster, Engineering-Heavy: What Owners Can Learn From AI Startups

AI is Birthing New Organization Forms

Firms that build AI into what they sell look different on the inside. A study by Hyunjin Kim of INSEAD and Rembrand Koning of Harvard Business School examined about 2,900 Y Combinator startups. Compared with peers in the same industry and cohort, the AI-embedded firms ran with roughly 25% fewer people and about half a level less hierarchy. Engineers and scientists made up 45% of their staff, against 36% at the others. Entry-level and management ranks were each thinner. Investors valued them comparably, which implies a much higher valuation per employee.

Read that as an early signal, not a verdict. These are young companies, and nobody knows whether the shape holds at 200 people. But the pattern is reasonable, and the reason behind it matters more than the numbers.

The Shape Follows the Cognitive Capital

These firms are shaped to reflect where their cognitive capital lives. When AI is built into the product, the customer operates the intelligence directly. The scoping, routing, reviewing and approving that once required people and managers has less to do. What remains is engineering: building the product, improving it and handling the cases it cannot.

That is the logic, and it does not depend on startups. It applies to any company that takes cognitive work and moves it into software.

The Incumbents Are Already Doing It

Look at the established information companies. S&P Global has been putting AI into its data and analytics products for years, through Kensho and the AI features around Capital IQ. Thomson Reuters is building CoCounsel into its legal, tax and research products. Both started with decades of proprietary content and expert judgment, and both are working to make that judgment something the customer can use without a human in between. I expect that they have a similar engineering heavy organization, flatter too.