Owner.com now routes 83% of new customers through a product it gives away free

Take the residual first. If more than 83% of new customers now begin inside an AI product ^[13], and the funnel was previously 100% sales-led inbound ^[3], then fewer than 17% of new customers meet a human before they meet the software ^[19]. That subtraction is the entire story, and note what it is not measuring: nothing in the account says the restaurant websites got better, only that acquiring one stopped requiring a demo booking, a salesperson and an onboarding specialist ^[2].

The part I find more interesting than Grader is the lead qualification agent, which estimates gross payments volume for a restaurant it has never worked with to within about $250, before anyone at Owner speaks to it ^[15]. Owner takes a cut of payment volume rather than a seat fee ^[7], so a pre-contact estimate of payment volume is a pre-contact estimate of lifetime revenue. That is underwriting. And it matters because the free tier is expensive: a finished website, upscaled photography, generated video and a full SEO and CRO audit are delivered before anyone pays ^[4], on top of a crawl that checks roughly 90 factors, pulls nearby competitors and reads the restaurant's reviews ^[9]. Free-for-everyone only pencils if you can rank prospects before you spend. Or rather, the more useful version: it makes free-for-everyone selectively affordable, which is a different business than a free trial.

This is where the copy-this framing gets thin. Guild's metric inversion, in which every login to fix what the software did counts as a failure of the software ^[5], is survivable because the money follows the restaurant's sales rather than the restaurant's clicking ^[7]. On flat per-seat pricing the same inversion bills you for the behaviour you just declared a defect, and the renewal conversation becomes an argument about shelfware ^[8]. Growth teams inherit the same problem from the other end, since most activation definitions are some version of returning within seven days and most retention cohorts are login cohorts ^[18]. So the transferable lesson is narrow: outcome-priced companies can afford to make their software invisible, seat-priced ones have to reprice first.

What would prove the distribution thesis wrong is conversion, and the source does not have it. There is an entry-mix number and an ARR level ^[13][14], and no CAC, no free-to-paid rate, no gross margin on the giveaway. If the Grader cohort converts materially worse than the demo-booked cohort did, Owner has bought volume with delivered cost of goods and the old funnel was cheaper per closed dollar. Second reading: growth came from restaurant payment volume expanding under a take rate ^[7], and the AI entry path is correlated rather than causal. Third: it works and holds, in which case the defensible asset is the open-web crawl and review corpus ^[9], not model access. Worth pricing the sourcing too, since these are stage numbers from the CEO at SaaStr AI 2026 written up by the investor who led the seed round and sits on the board ^[1], which is informed rather than audited.

The allocation point is the quiet one. Three years of build went into the path in and into internal coordination, where an agent named Owen absorbs about 90% of builder coordination and finance moved its primary artifact out of Excel into Claude ^[16], while the rebuild itself was elective at a company already exceeding its plan ^[17]. Guild shipped five products himself in two months without ever having written production code at Owner ^[6]. One every twelve days ^[12].