Cognitive warmup. A Pizza Hut franchise, which operates more than 110 Pizza Hut outlets in North America, is suing the restaurant giant for using AI that has slowed down once high-performing outlets, resulting in a mess of delays that has hurt sales, seen customer satisfaction plummet and generally been chaotic for restaurant operations. Chaac Pizza Northeast which operates stores across New York, New Jersey, Maryland, Pennsylvania and Washington, D.C., is seeking $100 million in damages for the chain’s Dragontail AI dispatch system.
The lawsuit states that the AI system gave DoorDash drivers unusual transparency into specific kitchen operations, allowing them to meddle with the system by delaying pickups, clubbing deliveries together, and selectively picking higher-tip orders. The franchise says that before this AI system, more than 90% of its pizza deliveries reached customers within 30 minutes. Our conversation this week has a larger theme. The cost and the value of intelligence. Many don’t realise the two are different.
“Don’t use AI just for the sake of using AI”
Yes, you read that right. “Don’t use AI just for the sake of using AI.” The words of Amazon executive Dave Treadwell, in an internal communication to employees that was reported widely last week. Probably the smartest thing I have heard an AI executive speak, in the longest time. But do you know why we reached this point? Because infected by enthusiasm, Amazon at some point instituted an internal leaderboard system which tracked employees on AI usage.
Cue, some workers began to assign AI agents (as you’d know from our previous conversations, these are autonomous AI bots that can do certain tasks automatically as defined by the user). Turns out, much of was a comprehensive collection of tasks that weren’t needed at all, or pointless in the grand scheme of things. The mission, to climb those rankings board which Amazon had implemented. This is what happens when an unnecessary sense of competitiveness is inserted into a workplace culture, to keep a narrative chugging along. Clearly, the costs of AI (which no one talked about in that excitement to replace humans), are significantly higher than actual humans being paid salaries.
AI bills are exorbitant
AI costs more than humans it wants to replace. Turns out, companies could no longer hide this reality. The big ones are the first to fall, in a way. Microsoft, in December, had begun to motivate its own developer teams to use more AI including Anthropic’s Claude Code, has now started to place revoke dates on those subscriptions for employees. Turns out, it was a popular tool and the costs were proving a bit much for Microsoft. They want their developers to use the Copilot CLI instead.
Here’s a minor detail—June 30th is the last day of Microsoft’s financial year. Cancelling Claude Code licenses at that cutoff helps cut some operating expenses, ahead of their new financial year.
They aren’t the only ones.
Uber is believed to have finished its entire budgetary allocation for AI in 2026, in just four months. “If you’re not actually able to draw a direct line to how (many) useful features and functionality you’re shipping to your users, that trade becomes harder to justify,” Uber president and chief operating officer Andrew Macdonald said in an interview. Uber had deployed AI agents for its engineers (believed to be around 5,000. It is becoming very hard to explain these AI usage bills, isn’t it?
Some of you may remember, a couple of weeks ago, we had talked about an Nvidia exec saying that “ the cost of compute is far beyond the costs of the employees”. I’d have loved to be a fly on the wall, the next time this exec bumped into Jensen Huang in the corridors of Nvidia’s headquarters. And now I wonder, how soon will the blame shift to humans, for using too much AI?
