700 engineers pretended to be AI. Investors paid ₹3,800cr for it

Builder.ai told customers its assistant Natasha could write app code on its own. It couldn't. For years, roughly 700 engineers in India sat behind the chatbot, manually writing the code and delaying delivery so it looked like AI was thinking. The company raised significant funding from investors including Microsoft and Qatar's sovereign wealth fund. It filed for bankruptcy in May 2025.

The Pizza Pitch That Hid the Real Workforce

Builder.ai's pitch was disarmingly simple: building an app should be as easy as ordering pizza. Pick your features from a menu, and an AI assistant named Natasha would generate the app. That story, plus a name-drop of Microsoft as an investor, helped the company raise significant funding.

The problem: Natasha wasn't writing the apps. Rest of World's reporting on the company describes staff working as "productologists" — a title that, in any other company, would just be product manager — coordinating human engineers who wrote the actual code by hand. The AI story was the product. The 700 engineers were the operation running underneath it, invisible to the customer and, allegedly, understated to investors.

This wasn't a slip. As early as 2019, the Wall Street Journal had already reported that Builder.ai leaned heavily on human contractors, with former employees describing it internally as "all engineer, no AI." A former employee reportedly sued the company that same year, alleging he was fired after flagging that the tech "did not work as promoted and was essentially nothing more than 'smoke and mirrors.'" Court filings claimed the company told investors apps were "80% built" by AI it had "barely even begun to develop." The lie had a six-year head start before it caught up with the company.

The 12-to-48-hour Lie That Made Slow Look Smart

Here's the detail that separates this from ordinary outsourcing: the deception wasn't just claiming humans were AI. It was engineering the delay to sell the illusion. Coverage of the collapse describes engineers being told to mimic AI response times by delaying code delivery by 12 to 48 hours — manufacturing a wait that made the output feel machine-generated rather than human-typed.

Think about what that means operationally. A genuine outsourcing shop in Pune or Bengaluru would ship a customer's code the moment it was ready — that's the entire value of hiring humans instead of waiting on a model to improve. Builder.ai's engineers were told to sit on finished work so it would look automated. Customers who thought they were getting AI speed were, by design, getting slower service than an honest outsourcing vendor would have given them. The fraud didn't just misrepresent the labor — it degraded the actual product to protect the story.

That's the part worth sitting with if you run any kind of dev shop, agency, or SaaS tool with a "smart" layer on top of manual work: the moment you start managing perception instead of managing delivery, your actual customers are the ones who pay the tax.

The Audit That Found the Revenue Gap

The collapse wasn't triggered by a whistleblower or a journalist. It was triggered by a lender running the numbers. In May 2025, Viola Credit — which had lent Builder.ai money — seized funds from the company's accounts after discovering the startup had significantly inflated its 2024 revenue projections.

Founder Sachin Dev Duggal had told creditors to expect substantially higher sales. An independent audit found actual revenue was significantly lower than promised. The gap between what was promised and what existed triggered the seizure of company accounts. New CEO Manpreet Ratia, brought in that February specifically to clean up the business, is the one who surfaced the extent of the misrepresentation once he actually opened the books.

This is the same pattern operators online have pointed at in a completely different context this year — big tech companies reporting headline profit numbers that don't survive a look at the footnotes, where the real story sits buried in a single line nobody reads past the summary.

Strip away the framing and it's the same lesson at every scale, from inflated sales forecasts to headline profit figures: a number that only survives if nobody checks the footnote isn't a real number. It's a placeholder for a story someone needs you to believe.

Why Microsoft and Qatar Didn't Catch It Sooner

The uncomfortable question in every version of this story: how does a company with Microsoft and a sovereign wealth fund on its cap table go six years — 2019 to 2025 — with public reporting, a lawsuit, and internal descriptions of "no AI" already circulating, before anyone with real money forces an audit?

Part of the answer is structural, and it isn't unique to Builder.ai. Big investors are underwriting a story, not auditing a codebase line by line. A brand name like Microsoft attached to a cap table functions as social proof for every investor who comes after — nobody wants to be the one asking uncomfortable questions when a name that big already wrote the check. That's a version of what founders online have been openly skeptical about all year in tech financial reporting generally: numbers that look strong on the surface can be propped up by whoever benefits most from them looking strong.

The other part is boring and more useful: nobody forced the books open until a lender's own money was at direct risk. Viola Credit didn't discover the inflation because it was virtuous — it discovered it because it had funds sitting in Builder.ai's accounts and a reason to check. That's the actual lesson for anyone raising money on a story: the people who eventually audit you aren't your investors. They're whoever's cash is exposed when the story stops matching the bank balance.

AI-washing Isn't New. It's Just Got a Name Now

Coverage of the Builder.ai collapse consistently uses one phrase: "AI-washing" — claiming AI capability you don't actually have, the same instinct as greenwashing a product's environmental impact. What makes Builder.ai the textbook case isn't that it happened once. It's that the company built an entire operational process — the 12-to-48-hour delay — specifically to make the wash convincing.

This matters right now because every SaaS founder pitching in 2025-26 has an incentive to say "AI-powered" somewhere on the homepage, whether or not there's a model doing meaningful work under the hood. The difference between reasonable positioning and Builder.ai-style fraud is simple: are you describing what the product does, or are you designing operational theater to hide what it doesn't do? A company that says "AI-assisted" when a human reviews every output is fine. A company that trains its staff to delay real human output so it feels automated has crossed into fabrication.

If you're a founder wondering whether your own "AI" claim would survive an audit, that's worth checking before an investor's lender does it for you — the same way a bootstrapped accounting startup built its ₹8.5 crore in revenue without needing to dress up its actual mechanism at all.

The Real Cost of Lying to Your Own Cap Table

The financial mechanics of Builder.ai's fraud eventually generated a US federal investigation, on top of the bankruptcy filing itself. That's the tail end most founders never think about when they're tempted to round a number up on a pitch deck: inflating revenue significantly isn't a growth-hacking move, it's a fraud claim waiting for the first person with the authority to demand an audit.

The pattern here is worth internalizing even outside outsourcing and AI: a founder who oversells a metric once buys themselves a short-term valuation bump and a long-term liability that compounds with every quarter the gap stays hidden. The promised revenue figures didn't just fail to materialize — the fact that they were promised at all is what triggered the seizure of company accounts, the ousting, and eventually the collapse of a company that, without the lie, might have simply been a smaller, honest outsourcing business that survived.

That's the actual trade every founder is making when they inflate a number for one more funding round: you're not buying time, you're buying a bigger crater for when the real number finally surfaces.

The Honest Version of "fake It Till You Make It"

There's a version of "fake it till you make it" that's genuinely normal in early-stage building — a landing page before the backend exists, a manual process standing in for automation you'll build once you have paying customers to justify it. The difference between that and Builder.ai is disclosure and direction.

A founder running a manual process while telling investors "we're automating this as we scale" is being honest about the gap. Builder.ai told investors and customers the gap didn't exist, then built a delay mechanism to make the illusion durable under scrutiny. One is a startup finding its footing. The other is a company engineering evidence for a lie.

Deepak Subramanian, Founder & CEO of YourTribe, watched Builder.ai work the startup-event circuit for years before the collapse and drew the same conclusion from the other side of the booth:

That's the same six-year gap in one sentence: Builder.ai had the booths, the Microsoft name-drop, and the Davos speaking slots. It never had the fundamentals underneath them.

If there's a single operational check every founder claiming any AI capability should run today, it's this: open your own product and ask what a skeptical auditor, not a friendly investor, would find in the first hour. Builder.ai's whole story survived because nobody with enough leverage asked that question until a lender's money was already on the line.

Conclusion

Builder.ai didn't collapse because it used human engineers — plenty of honest outsourcing businesses do that profitably. It collapsed because it built a delay mechanism to disguise labor as intelligence, then inflated revenue to match the story. If any part of your own pitch depends on nobody opening the footnote, open it yourself today, before a lender does.