The AI Startup That Was Just Humans in a Trench Coat

Remember when every startup had to slap “AI” on its pitch deck to get funded? Engineer.ai was the poster child for that era. The Indian startup raised nearly $30 million from SoftBank and others, claiming it had built an AI platform that could automate 80% of mobile app development in about an hour. Its founder, Sachin Dev Duggal, who apparently goes by “Chief Wizard” (I’m not making that up), even took the stage at a conference to sell this vision.

Turns out the only magic happening was good old-fashioned human labor. A Wall Street Journal investigation revealed that Engineer.ai wasn’t using AI to assemble code at all. It was using human engineers in India and elsewhere to build those apps. The AI was basically a marketing gimmick.

This isn’t just a case of overpromising. The company was sued earlier this year by its own chief business officer, Robert Holdheim, who claimed Duggal was telling investors the platform was 80% done when he’d barely started. That’s not a stretch goal — that’s a lie.

When pressed on what AI they actually used, Engineer.ai pointed to natural language processing for estimating pricing and timelines, and a “decision tree” for assigning tasks to engineers. Neither of those qualifies as the kind of modern AI that powers self-driving cars or image recognition. No AI was compiling code. No machine learning was generating features. It was just a project management tool with a fancy coat of paint.

What’s depressing is how common this is. A study by MMC Ventures found that startups claiming an AI component can attract 50% more funding than other software companies. They also estimated that 40% or more of those companies don’t use any real AI at all. The .ai domain from Anguilla has doubled in recent years. Everyone wants a piece of the hype.

The problem is that real AI is hard. Getting training data, building models, deploying at scale — that takes years and millions of dollars. Companies like Facebook and Google have entire research orgs dedicated to it. But a startup can just say “AI” and watch the money roll in, as long as they can keep the charade going long enough to actually build something.

Engineer.ai isn’t an anomaly. It’s a symptom of an industry where buzzwords matter more than substance. The uncomfortable truth is that a lot of “AI” today barely exists. Content moderation on Facebook and YouTube? Some AI, but mostly armies of contractors reviewing content manually. Customer service chatbots? Often humans pretending to be bots until the AI fails.

AI washing is a real problem, and it’s not going away. The incentives are all wrong. Investors want to hear about AI. Customers want to believe in automation. And startups need funding to survive. So they take shortcuts. They exaggerate. They hire humans and call it AI.

But here’s the thing: this approach has been tried before. Remember the Theranos scandal? The fake it till you make it strategy usually ends the same way. Engineer.ai might have gotten away with it for a while, but the truth always comes out. And when it does, the trust evaporates.

I don’t blame the engineers at Engineer.ai for doing the work. They’re just trying to make a living. But the founders and investors who knew the truth and kept selling the dream? That’s a different story. The AI hype cycle has created a lot of noise, but it’s also created a lot of damage. Real progress gets drowned out by the noise, and genuine companies suffer when the bubble bursts.

So next time you see a startup claiming AI can do the impossible, ask for the receipts. Ask how the model was trained. Ask for a demo. And if they start talking about “human-assisted” AI, you know what’s really going on.