As technology gets bigger, its failures get bigger too — and sometimes so do the efforts to hide them. For example, recently, a wave of stories has exposed ‘AI’ products that were really human‑powered behind the scenes.
A prominent example involves a London-based company, Builder.AI, which at one point was valued at $1.5 billion, that was exposed for secretly employing approximately 700 real people to perform services it marketed as AI-delivered. This company, which has since filed for bankruptcy, received investments from major firms including Microsoft.
Other reports have highlighted similar patterns:
- A company providing “AI-powered” voice interfaces for fast-food drive-thrus could only complete 30% of orders without human intervention.
- Amazon was found to have secretly relied on real employees while promoting an “AI” product.
- NEO, the home robot, was marketed as a butler that could perform any of your chores reliably … but took two minutes to fold a sweater, couldn’t crack a walnut, and was teleoperated the entire time.
These incidents demonstrate a pattern of companies leveraging AI hype to win investment and customers, while hiding how much work is still done by humans.
There’s nothing wrong with humans in the loop; the problem is pretending they aren’t there and selling that pretense as innovation.
But hiding humans wasn’t the only way tech disappointed us this year.
Finding New Ways to Fail
A recent ChatGPT update was sycophantic to a fault, assuring users that even their most mundane ideas were brilliant and incisive. Unfortunately, OpenAI responded by swinging the pendulum too far in the other direction. Their next update, GPT-5, was so cold that it prompted them to revive the ability to choose which model you used, and likely contributed to Altman’s recent “code red”.
Or, you can point to the countless “meme coins” that made money only for their creators before being rug-pulled, such as the Hawk Tuah Coin.
The Path To Success
The common thread isn’t that technology is moving too fast — it’s that too many people are trying to leap over the boring parts.
Many of this year’s failures were caused by people trying to skip the fundamentals.
Meme coins didn’t fail because communities don’t matter — they failed because speculation was mistaken for value. The humanoid robots didn’t disappoint because robotics is a dead end — they disappointed because demos were sold as deployments. And the companies quietly swapping humans in for “AI” didn’t collapse because AI is useless — they collapsed because trust, once broken, is almost impossible to recover.
A What Not To Do List
These truths sound obvious, but the past year suggests many leaders still ignore them.
- Don’t hide humans and call it AI.
- Don’t sell demos as finished products.
- Don’t mistake speculation for sustainable value.
- Don’t optimize for virality at the expense of trust.
For entrepreneurs, the lesson is uncomfortable but simple: reality wins in the long run. You can borrow attention for a moment, but you have to earn durability. Markets and customers will forgive slow progress, but they won’t forgive dishonesty.
What To Do Instead
- Validate in the real world.
- Disclose human‑in‑the‑loop honestly.
- Align metrics with durability.
- Design for boring reliability before spectacle.
In some ways, it’s easier than ever to ‘succeed’. With that said, does success simply mean building something that works, or does it mean building something that’s strategic and unique and captures the imagination and wallets of an audience big enough to fuel your desired bigger future?
It’s the same paradox that AI‑created marketing faces. It’s now much easier to create something that sounds logical, but it is harder to stand out because you’re competing for attention in a growing sea of sameness and noise.
The next generation of meaningful companies won’t be built by chasing the loudest narrative or the newest acronym. They’ll be built by founders who understand the difference between a prototype and a product, between timely and timeless, and between promise and proof.
Hype can open the door. Execution keeps it open.
Leave a Reply