They should eliminate about 60% of the menu and focus on speed and quality for the remainder.
That's the idea. Part of what they're trying to do is speed up drive-thru service, since that's now where the majority of customers visit.
Humans can't predict what to make in the minutes before a customer pulls up to the speaker, but an ML tied into a number of systems and inputs might be able to do a better job.
It's not really anything new, to be honest. Back in the days where restaurants cooked and wrapped the burgers ahead of time and held them in a warming bin for 10 minutes at a time (at least that's what the spec dictated) the managers would keep detailed records on customer count, weather, cars in the lot, local events, etc and try to make predictions based on historical data. Wasn't the most effective, but it helped. This is just a modernization of that idea.
MCD will most likely spend half a billion dollars to get that burger on the grill 90 seconds earlier if it's possible.
Mcdonalds has a mobile app, where you can order and your food will be ready before you arrive. Wouldn't it have been better to push customers towards this channel, where they can order what they want instead of ML guessing, and saving a couple hundred million dollars along the way?
As others have mentioned, Dominoes doesn't use ML/AI, but you can order a pizza crazy fast through the app and know when it's done (and that seems to be enough).
From my interactions with their people, the mobile app is the conduit to the future but there are a lot of kinks still left to be worked out.
For example: you order on the app, but when exactly do you start the food? You might be 30 seconds from the store or 30 minutes. The initial idea was to geofence the store and trigger the kitchen when you're on the way. Perhaps that's still going, but maybe the ML is an assist on that data. You also get a ton of marketing and demographic information out of that. There are also the usual privacy concerns.
Remember that there's also a ton of data being generated in the other direction. Predictive analytics on the equipment (everything in the kitchen will be an edge node in a few years), POS/cash accounting, real-time sales data, store traffic, crew work history, etc.
They're also experimenting with dynamic pricing on food, like it or not.
Small rant: a weird pleasure in Japan is using restaurant touchscreen consoles that don't take several million very laggy taps to complete a simple order. Places like Yayoiken[1] and Matsuya[2] have this down to a science. It's partly because of fewer menu options (although it's not like either one has a miniscule menu[3]).
Insert cash, tap tile, ticket immediately dispenses. Tap change button when done. The combination of denser screen layouts, decently responsive touchscreens and a near-complete lack of interstitial animations makes using these things extremely snappy. If you know what you want, the entire process can take maybe ten seconds.
[1] https://s3-media1.fl.yelpcdn.com/bphoto/PzkepdRwWT6NaE4XDgKz...
[2] http://3.bp.blogspot.com/-cJ8-6hscmOQ/ViHzR4f0EmI/AAAAAAAABc...
If I know facebook, they are going to make McDonalds addicting AF