The pitch — name stripped
A mobile app that lets anyone buy any product from any online store with a single tap — no logins, no carts, no checkout forms. Pitched as proprietary AI that completes the whole purchase end to end, automatically, across every merchant on the internet, with no integrations and a claimed automation rate north of 90 percent — all sold to investors before there is proof the AI can reliably finish even one checkout on a site it has never seen.
Kasspian’s cold read on Nate
Fatal flawThat a single AI agent could reliably complete checkout on any merchant on the internet — thousands of unique, constantly changing, anti-bot-defended flows — with no integrations and no human stepping in, and that this was already solved at 90-plus percent.
The company launched in 2018 with a clean promise: buy from any e-commerce site with one tap. No account on the merchant, no shipping form, no card entry — the app's AI would handle the entire purchase for you. The founder told investors, across rounds from 2019 through 2022, that the app could transact online 'without human intervention,' and put the automation rate at 93 to 97 percent. That number was the whole story. A universal checkout that runs itself is a software company with software margins. A universal checkout that needs a person is a staffing agency.
The capital followed the number. The company raised over $50 million — roughly $51 million in equity — anchored by a $38 million Series A in June 2021 led by Renegade Partners, with Coatue, Forerunner Ventures, and Canaan Partners alongside. The pitch was a frictionless commerce layer for the entire internet, and the AI claim was what made the multiple make sense. Nobody pays a software multiple for a call center.
The reality was the inverse of the pitch. The founder had bought AI technology from a third party and hired data scientists, but the system never learned to reliably complete purchases. Prosecutors later put the app's actual automation rate at 'effectively zero percent.' The orders were finished by hundreds of contractors in a call center in the Philippines — internally called purchasing assistants — manually placing each one by hand. The human work was not the rare edge case the founder described. It was the product.
The economics did what they always do when the engine is people, not code: they ran out. The company burned through its cash and was forced to sell its assets in January 2023, leaving investors with what the indictment called near-total losses. In April 2025 the U.S. Attorney for the Southern District of New York charged the founder, then 35, with one count of securities fraud and one count of wire fraud, each carrying up to 20 years. The SEC filed parallel civil charges, and regulators framed the case as a warning shot on 'AI washing.'
The fraud is the headline, but the structural tell was visible on day one, before any lie was needed. If a human in a call center completes every order, your cost per order does not fall as you grow — it tracks headcount, order for order, forever. The AI label was not a feature. It was a tarp thrown over a payroll line that scaled linearly with revenue. The crime was lying about the automation rate. The business mistake was needing the automation rate to be true and never proving it was.
Strip the name off this pitch and the riskiest assumption is sitting in the open: that one AI agent can complete checkout on any merchant on earth, with no integrations, and that it already works more than 90 percent of the time. That is the single hardest claim in the deck, and it is stated as solved. Kasspian would flag it on sight — when the most technically improbable part of a pitch is presented as the part you no longer have to worry about, that is exactly where to put the diligence. The headline automation rate was not a metric. It was the entire valuation.
The premise was broken in the math, not just the morals. A universal checkout that the AI cannot finish has to be finished by a person — so every order the model misses becomes a human order, paid by the hour. At a real automation rate near zero, every single order was a human order. That is not a software company with a margin that improves as it scales; it is a labor company whose costs rise in lockstep with revenue, dressed in a software multiple. Capital can buy reach, it can't buy a unit that makes money — and here each new tap of growth simply bought another seat in the call center.
The cheap test was one audited number, available years before the Series A. What percentage of orders complete with no human touching them? Pull a random week of transactions, open the logs, count the ones a person finished. No diligence team required — just the data the company already had. Prove the automation rate before you raise on the automation rate. A flawed premise doesn't get less flawed with more capital; in this case the money just paid for a bigger room of people pretending to be the machine.
Why did Nate fail?
That a single AI agent could reliably complete checkout on any merchant on the internet — thousands of unique, constantly changing, anti-bot-defended flows — with no integrations and no human stepping in, and that this was already solved at 90-plus percent.
What actually happened to Nate?
The company launched in 2018 with a clean promise: buy from any e-commerce site with one tap. No account on the merchant, no shipping form, no card entry — the app's AI would handle the entire purchase for you. The founder told investors, across rounds from 2019 through 2022, that the app could transact online 'without human intervention,' and put the automation rate at 93 to 97 percent. That number was the whole story. A universal checkout that runs itself is a software company with software margins. A universal checkout that needs a person is a staffing agency.
What can founders learn from Nate?
Strip the name off this pitch and the riskiest assumption is sitting in the open: that one AI agent can complete checkout on any merchant on earth, with no integrations, and that it already works more than 90 percent of the time. That is the single hardest claim in the deck, and it is stated as solved. Kasspian would flag it on sight — when the most technically improbable part of a pitch is presented as the part you no longer have to worry about, that is exactly where to put the diligence. The headline automation rate was not a metric. It was the entire valuation.
Nate looked like a good idea too. Get the same honest read on yours — score, fatal flaw, market — in about 90 seconds.
One week, a dead startup and why it died. The next, a play to get your next customers.
No fluff, no spam. Unsubscribe anytime.