AI Startups Could Be Worth Payroll, Not Just Software
A law firm’s software tab is small, but its paralegal payroll is huge, and that difference could decide what AI companies are really worth.

Invest Like the Best with Patrick O'Shaughnessy · Ramtin Naimi on the State of Venture Investing - [Invest Like the Best, EP.499]Public software companies trade at about 13 times revenue, yet some vertical software businesses are getting 100 times revenue. Venture investor Ramtin Naimi says the gap turns on a simple question: are these software businesses or labor replacement businesses?
If they are software businesses, the high prices look hard to justify. A customer like a law firm spends only a limited amount on software each year, which limits how large any supplier can become.
If they are labor businesses, the ceiling is much higher. The same law firms spend far more on paralegals each year than they do on software, and that payroll spending is the larger market an AI company would be chasing.
Skeptics do not buy the labor replacement idea, Naimi says, in part because he has not yet seen that replacement actually happen. What keeps believers interested is the pace of growth, with some of these companies still growing five, seven, or ten times year over year even after reaching massive scale.
He says companies in his portfolio are expected to increase by 200 to 700 over this year while their margins move from negative 15 percent to positive 35 percent. That is a forecast for his portfolio, not a result, with rapid revenue growth and improving profitability happening together.
So the valuation test is which budget the startup can tap. Measured against software budgets, such a company looks small and expensive. Measured against payroll, it can look much larger, if machines start doing work that people used to do.
Questions people ask
Why would replacing workers be worth more than selling software?
The reasoning follows Naimi's law firm example. Law firms spend relatively little on software each year, but much more on paralegals. If an AI product can do paralegal work, it is competing for that much larger payroll budget rather than the smaller software budget.
Has AI actually started replacing workers?
Naimi says he has not yet seen broad labor replacement happen. Many AI tools are still helpers that make employees faster, rather than substitutes that let employers eliminate roles, so the payroll market remains mostly potential.
What could go wrong with paying 100 times revenue?
Growth could slow, margins could stall, or customers could treat the product as just another tool and pay software prices for it. Naimi's benchmark for that software-only case is the public software market at about 13 times revenue, far below the 100 times seen for some startups.
