Either you or someone you know has probably worked at McDonald’s at some point in their lives. For most people, it’s working the drive-thru section while in high school or going to college. And it’s also long been a place where those older can go to be employed due to financial or career disruptions. Other than other fast-food restaurants, McDonald’s has even been a dependable place to be hired without having the fear of being replaced by technology. For a while, it kind of looked like that would sustain after failed attempts at using AI tech at select McDonald’s locations.
I’m talking about their Archy AI bot, something created to replace/supplement drive-thru attendants and supposedly be more accurate and/or efficient. It turns out, because it was a technological failure from a once powerful computing company, Mickey D’s is now rebooting the entire process.
What happens next is anyone’s guess considering AI is still very unpredictable in what it can do when left to its own devices. McDonald’s insists, though, that things will be different this time compared to a mere two years ago when customers became annoyed at Archy’s low IQ.
This is a Test—of the Archy Bot System
Just a few years ago, McDonald’s began using the above Archy bot in many of their drive-thrus throughout America. Whether you experienced that in your city was completely random, but there was reported use in at least 100 or more American locations. That’s a relatively small number, yet you may find someone who remembers what it was like ordering from Archy.
It all started in 2021 when AI was still in its infancy compared to now. IBM was the company tapped to create Archy, which seemed like a logical move at the time considering IBM’s longstanding computing legacy. However, everyone soon learned IBM was perhaps not going to be the leader in AI technology based on the performance of Archy in those above restaurants. Reports were it kept screwing up orders and didn’t completely understand conversations of customers. Apparently it was a lot like prior human drive-thru speakers where you (and drive-thru attendants) could barely be understood.
This ended up with countless order mistakes, including double orders for some people. You may remember seeing reports of this on social media where people reported just how bad it was. Somehow, McDonald’s kept it all going until the summer of 2024 when they finally abandoned the concept.
They looked at it like a learning experience, though, rather than a massive technological failure. There was a recent announcement they’ll be testing a new and improved Archy bot (this time from Google and rebranded ArchyIQ) in only 5 select locations nationwide. If it performs well, they’ll expand it—likely to your own town or city.
Therein lies the problem coming from Google’s Gemini, and namely all AI models. Almost all of them still get confused on certain questions, or when there’s multiple problems to solve. A McDonald’s drive-thru is already a recurring hotbed of confusion from most occupants in a car.
What Could Possibly Go Wrong?
The brighter side of the new ArchyIQ bot is that it’s built to understand conversational speech, including comprehending multiple languages. All of this goes on the assumption humans can literally hold a normal conversation or talk legibly in their own language when ordering in the drive-thru lane. Humans are a lot more unpredictable than the AI bots are, including how they behave when ordering food. You know this yourself if you have a family and try to order in a drive-thru system while everyone’s talking at once.
There’s a lot of chaos when it comes to families ordering food this way, and you have to wonder if ArchyIQ is really built to overcome that. Plus, it’s worth pondering whether “conversational speech” really means talking as legibly as possible so the bot can process what’s said. I know far too many people who kind of mumble when they talk, including when they’re required to speak up.
We’ll know soon enough whether this truly works, or the human race will completely confuse yet another AI bot with our idiosyncrasies. Even if some will say it’s subtle revenge for AI bots confusing us with some of its own weirdness. My guess is it’s still going to need some tweaking so Archy can truly decipher some of the car chaos of families ordering food. Otherwise, it may be another AI bot receiving a bunch of verbal abuse for not getting something right.
It’ll Still Help Relieve the Stress of Human Workers
McDonald’s is pretty much insinuating the old restaurant idiom of “Do you want it good, or fast?” We’ve all seen how quality can be compromised when the humans hired at busy McDonald’s drive-thrus become frazzled from peak-hour chaos. I’ve experienced more times than I can count when orders got screwed up due to an employee trying to juggle multiple orders at once. In some cases, it was receiving someone else’s order by mistake, an item missing—or an inadvertent extra item.
As efficient as AI bots are supposed to be, the confusion of not understanding the quirks of human speech could end up doing the same thing repeatedly. Then again, the intention is to help the human workers have more free time to properly prepare the food and not do any screw-ups either. It may be a better bet where any mistake made won’t fall on the humans and just the Archy bot.
The good news is McDonald’s is trying to settle the argument whether humans should be done away with in favor of drive-thru bots, or have both work together. They seem to be trying to set a precedent of making drive-thru jobs less stressful and more appealing. And for those who often have aversions to the idea of having to work at McDonald’s, it may offer more peace of mind when a time comes one has to work there for experience or loss of other job.
Or, will AI turn on the human workers and try to make a screw-up look like it was the latter’s fault instead of itself? That’s not necessarily something to dismiss in a world where many companies are already using AI systems to track work performance of employees. The final word goes to the bot in many cases, making bots almost like a judge and jury on certain errors made.
A more hopeful scenario is that something like ArchyIQ would fall on its own grenade if numerous faux pas ordering scenarios start mounting this year.
Will Google Be Better Than IBM?
If you’ve used Google’s AI system on their search engines, you know it’s not completely fail safe on providing information. While Gemini is used by many millions now for more advanced AI roles, there isn’t any system in this realm that’s completely error-free perfect. It seems some mistakes are going to be made, with maybe a lesser error rate than IBM’s several years ago.
According to some preliminary stats on McDonald’s ArchyIQ, one million test orders have been placed recently. Out of those, there was a 90% success rate on orders. While this is just preliminary data and not detailed yet, you have to do the math on that based on five test McDonald’s at the moment.
Having a 10% error rate in five locations could still be considered statistically more than McDonald’s likely wanted. Part of this may be based on the option available for humans to request a human when ordering at the drive-thru if they don’t want to order from ArchyIQ.
It opens the question of whether more people are ditching the idea of AI there and opting in to the old method of a human being. As McDonald’s starts expanding ArchyIQ, the choice of flipping to a human may be all the more obvious when looking at the numbers.
If that happens, it kind of gives hints to a current (and maybe temporary) backlash against AI taking over everything. I’ll be covering this societal aspect again in this blog, but the next few years may see a lot more choice to go human when given an option.
Also, there’s a new option for human employees to step in and intervene should ArchyIQ trip up on a particular order. You can call these the “AI watchers”, which is a new role developing in other companies as well. It’s humans watching the AI to step in and turn them off if they do something stupid. Ironic it is, though, that the same AI is probably watching the humans while we watch them. All of that comes down to a potential mental chess match.
Whether that happens at McDonald’s is something to watch. With so many AI systems escaping their confines and going rogue, imagine what could happen if ArchyIQ suddenly starts spewing lies in reporting human employees for something that potentially didn’t happen. Especially when a human has to turn it off with the claim the bot failed.
I’d try ArchyIQ out of curiosity to see if it gets my order right. With this all being a part of McDonald’s technological strategy of “McDonald’s Next”, what’s next may be just a growing demand to bring back the humans for the most part to rebirth McD’s old (and lost) character. Yeah, including the bored teens who took your orders during the summer season.
In Part 10, I’ll look at my experience using Suno Pro recently to create a studio quality recording for a song I wrote. It changed how I view AI as a creative assistant—without becoming complacent to letting it create everything.
/End