Will AI Replace Humans? Should It?
Aside from a childhood paper route, my first foray into the workforce was as a cashier at a McDonald’s here in Utah. I was a cashier, starting out on the front counter. Over time, I was assigned more and more to the drive-through where you need to multi-task more to take orders, handle money, and prepare drinks. It’s higher-pressure than the front counter cashiering and requires a more independent-minded worker.
As a former McD’s cashier and the co-founder of an AI company, I was interested to read the reports of the failure of McDonald’s foray into AI order-taking at their drive-throughs. The AI chatbots they used sometimes had difficulty getting orders right, doing things like adding 10,000 chicken Mcnuggets to an order. Eventually, the restaurant pulled the plug on the project, declaring the technology not mature enough to deploy more widely. They wanted the AI to replace their drive-through cashiers and run autonomously. That gets to a widespread fallacy in the discussions of the future of work in an AI world. Specifically, you see articles written in one of two extremes: either AI (and robotics and whatever) is going to put nearly everyone out of work, or AI is completely useless because it can’t actually put nearly everyone out of work.
Here’s the thing: McDonald’s could have designed their AI experience in a way that gave the tool a chance to be useful, even if it wasn’t perfect. It could have made things easier for their workers without eliminating their jobs. The choice shouldn’t be AI or humans in a zero-sum deathmatch fighting for world domination. The choice is AI and humans, working together to be more productive and to take the drudgery and stress out of our everyday jobs.
When I worked there (something more than twenty years ago), the drive-through staff varied between one and three people. At slow times, one person took orders, took payment, made drinks, and handed out the food. When it got busier, you’d split into one person taking orders and payment and one or two making drinks, getting ice cream (when the machine wasn’t broken, am I right?), and handing things out. The intent of the AI order-taker was apparently to eliminate at least two of the regular drive-through tasks.
But how much harder would it have been to have the worker making drinks still wear a headset and listen to the order? To fix any mistakes or garbled language or to ask for clarification if neither they nor the machine understood the mumbler with the deafening diesel engine? What if they had the backup to let the AI be accurate 90% of the time instead of 100%? Still way easier for the worker in the drive-through. It would free them up for additional tasks. Maybe they would shuffle things around and have the drive-through dropping the fries so the front counter can provide better service during the lunch rush. Or maybe someone can dart out into the dining room once in a while to wipe the ketchup off the tables before it gets all gummy and crusty.
Here at Oak Canyon, we’re not building tools to replace you or anyone else. We know the limitations of generative AI. We know that while it might be better at math than I am, it’s not as smart as you or me. But, if it can be good enough and easy enough to simplify your workflow and make your life easier? That’s worth having, even if it sometimes suggests adding glue to your pizza sauce. Because we both know you wouldn’t let that kind of mistake slide.