Engineer.ai’s ‘AI’ was just people in India — and we all saw it coming

Remember Engineer.ai? The Indian startup that promised to build 80% of your mobile app in an hour using AI? Turns out, that “AI” was just a bunch of human engineers in India and elsewhere, stitching code together by hand. The Wall Street Journal broke the story back in 2019, and honestly, it’s still one of the best examples of how hollow the AI hype machine can be.

The company had raised nearly $30 million from SoftBank’s venture arm and others. Its founder, Sachin Dev Duggal — who apparently also goes by “Chief Wizard” — was out there on stage claiming their AI platform could turn a vague idea into a working app in 60 minutes. But when you peeled back the layers, there was no AI writing code. No machine learning model compiling features. Just good old-fashioned software engineers doing the work, while the company pretended otherwise.

What makes this story stick is that it’s not an isolated case. MMC Ventures, a UK investment firm, found that startups with an AI label attract up to 50% more funding than those without. And they estimated that 40% or more of those companies don’t actually use any real AI. Engineer.ai was just the most blatant example.

The company’s own chief business officer, Robert Holdheim, sued them earlier that year. He claimed Duggal was telling investors the product was 80% done when it barely existed. That’s a pretty bold move when your title includes “Wizard.”

When pressed by the WSJ, Engineer.ai admitted they used natural language processing to estimate pricing and timelines, and a “decision tree” to assign tasks. That’s not modern AI. That’s barely a flowchart. No AI was assembling code. No neural network was learning from user requests. It was humans, plain and simple.

This whole saga reveals an uncomfortable truth about a lot of AI: it barely exists. We’ve seen this pattern before. Facebook and YouTube use some AI for moderation, but they also rely on armies of contractors to review content. Self-driving car companies have humans in the loop. “AI” often means “people we don’t have to call employees.”

Engineer.ai wasn’t alone in this game. The number of companies using the .ai domain doubled in just a few years, according to the WSJ. Slapping “AI” on your pitch deck became the easiest way to get meetings with VCs who were desperate to back the next big thing. And with SoftBank throwing hundreds of billions at AI, the incentive to exaggerate was huge.

The funny thing is, the company might have actually been trying to build real AI eventually. The funding was supposed to get them there. But they burned through trust first. And once you’re caught lying about your core technology, it’s hard to recover.

This story is a reminder that AI is hard. Really hard. Getting the training data, building the models, deploying at scale — that’s years of work, even for companies like Google and Facebook. Engineer.ai tried to skip the line by faking it. They got caught, and the industry got a little wiser for it.

So next time you hear a startup claim their AI can do something miraculous in an hour, ask yourself: where are the humans? Because they’re probably there, just not in the press release.