You’ve felt the pressure. The board wants an ‘AI strategy.’ Your competitors are launching agents. So, you mandate AI adoption across your entire company, expecting a productivity revolution.
What you usually get is chaos.
Look at what happened to Meituan, the massive Chinese food delivery giant known for surviving brutal price wars through ruthless, penny-pinching discipline. In early 2026, this famously frugal company did the unthinkable. They launched an internal ‘shrimp-raising’ campaign—forcing 100,000 employees to build AI agents. No token limits. No guardrails. Just the most expensive models on the market, handed to everyone from the C-suite to the frontline.
For two months, they burned millions a day on compute costs. But the financial hemorrhage wasn’t the real near-catastrophe.
The fear of being left behind is a much more powerful drug than the desire to lead.
When a company famous for counting pennies suddenly burns millions, you know they aren’t chasing a trend—they are running from a ghost. The ghost of missing the next platform shift. But this panic triggered what Meituan’s leadership later diagnosed as ‘firing before aiming.’
Employees didn’t know how to use the tools. They fed top-tier AI models with trivial tasks like drafting emails. Worst of all, they trusted the outputs. AI hallucinations—confident, fabricated nonsense—began leaking out of chat windows and directly into Meituan’s operational core.
AI hallucinating in a chatbot is a minor inconvenience. AI hallucinating in a delivery driver’s route or a merchant’s promotion rules is a catastrophic breach of trust.
When a senior executive stood up and admitted the experiment was ‘interfering with real operations’ and causing ‘security risks,’ the industry realized the truth: unleashing AI without organizational alignment isn’t innovation. It’s corporate self-sabotage.
Most commentators looked at Meituan’s burned millions and called it a colossal failure. They missed the point entirely.
Meituan didn’t fail. They bought an organizational vaccine.
By letting 100,000 employees touch the hot stove, leadership proved to the entire company that AI isn’t magic. You don’t buy transformation by issuing a memo and turning on an API. Real AI adoption requires business processes, organizational structures, and technology to evolve simultaneously.
After the ‘shrimp-raising’ disaster, Meituan pivoted. They stopped the free-for-all. They established dedicated AI divisions, launched internal competitions to find what actually worked, and built a structured, secure internal platform.
You don’t build an AI-first company by forcing everyone to use AI. You build it by forcing the organization to align around specific, high-stakes problems.
If you are a leader pushing AI inside a large organization right now, you are facing the exact same trap. The pressure to ‘do something with AI’ is suffocating. But broad, unfocused enthusiasm is just as dangerous as doing nothing.
Stop forcing your team to ‘use AI.’ Start forcing them to solve real business problems with AI. The distinction is the difference between burning millions on a corporate fever dream and building actual operational immunity.
FAQ
Q: Wasn't this just a massive waste of shareholder money?
A: No. It was an expensive calibration. By letting employees fail at scale, Meituan prevented much larger, systemic trust disasters down the line. It bought organizational realism.
Q: How do we adopt AI without burning millions?
A: Stop mandating broad AI usage. Identify specific, high-stakes operational bottlenecks, and align your tech, business processes, and org structure around solving them before you scale.
Q: Is the 'fear of missing out' actually a bad thing?
A: FOMO is a terrible master but an excellent accelerant. It forces action, but if you don't channel that panic into targeted execution, you'll just burn cash and breed cynicism.