Harvey, built for legal work, reached roughly $190 million in annual recurring revenue and an $11 billion valuation in March 2026, with more than 100,000 lawyers across 60 countries. Sierra, built for customer service, reached $100 million ARR in seven quarters while charging per resolved ticket.
Those are funded companies and their numbers are not a template for anyone reading this. The instructive part is what they chose not to build.
Neither one is general
Both had the option to be horizontal. The underlying capability does not care whether it reviews a lease or a support ticket, and a general product has a larger addressable market on paper.
They went narrow anyway, because the value is not in the capability. It is in knowing what correct looks like in one domain, and that knowledge does not generalise. What counts as a correctly resolved support ticket has nothing in common with what counts as a correctly reviewed contract, and no amount of model quality bridges that.
The smaller version of the same play
“The value is not in the capability. It is in knowing what correct looks like in one domain, and that knowledge does not generalise.”
Solo founders are routinely reaching $300,000 to $500,000 in annual recurring revenue on a single vertical agent, with small teams clearing a million. Those numbers matter more to most people reading this than Harvey's do.
At that scale the economics are unusual. A single-workflow product serving a few hundred customers in one industry needs no sales team, no category creation, and no funding. It needs someone who knows the workflow well enough to define correct.
The value is not in the capability.
Why this favours experience over technical skill
Turn what you know into what you own.
Vibepreneur builds structured ventures from professional expertise, with positioning, launch assets, and growth systems included.
Join the WaitlistDefining correct is the hard part and it is the part that cannot be researched. It requires having been wrong about it before, having seen the edge case that broke the rule, and knowing which exceptions matter and which are noise.
A technically strong founder entering an unfamiliar industry has to acquire that by observation, slowly, while shipping a product that is wrong in ways they cannot see. Someone who spent nine years in the industry starts with it and has to acquire the technical part instead, which is now the cheaper of the two.
That inversion is recent. For most of the last two decades it ran the other way.
What narrow actually requires
Not a small market. A specific one. The distinction matters because founders often narrow by shrinking their ambition rather than by sharpening their definition.
Sharpening means being able to state, in one sentence, the unit of work your product completes and the standard it must meet. If you cannot say that sentence yet, more building will not produce it. Talking to people who do the work will.
See the micro-specialist advantage for the services version of this, and how to find a vertical worth building in for choosing the domain.