AI and Drive-Thrus Make a Bad Combo
IF ANY OF YOU have been on TikTok lately (don’t judge me; it’s research) you might have seen people posting about their experiences with new AI-powered drive-thrus. My initial reaction: an eye roll so powerful it might have recalibrated the rotation of the earth.
All right, fine, perhaps a bit hyperbolic. But really?
I guess I shouldn’t be too surprised. After all, quick-service retail companies have been looking for ways to cut labor costs and reduce drive-thru wait times for years. Self-serve ordering kiosks and order-ahead apps have been pretty effective for reducing reliance on humans to take orders at the counter and to keep the lines moving.
Something that might be a little less obvious to the average patron is the trend of outsourced drive-thru operators. At least as far back as 2006, some of the larger fast-food chains have been running their drive-thru operations out of centralized centers supporting 40 or more locations. In the beginning of the pandemic, drive-thru volume kicked into high gear. We saw the emergence of vendors supplying restaurants with “virtual” drive-thru gig workers who can take orders for various restaurants from wherever they happen to be.
There is some sense to this—having dedicated drive-thru operators (versus folks who are going to be pulling double or even triple duty in the restaurant) allows for a focus on service quality and accuracy.
But back to the original issue of shifting to AI-powered drive-thrus. I’ll say it again: really?!
Speech recognition software has its challenges at the best of times. Many of us have the experience of calling a company’s 1-800 number and getting frustrated when the “phone robot” doesn’t understand what we’re saying, and conversations with the IVR tend to be pretty simple! You might get “there’s a mistake on my bill” or “I want to change my plan” and certainly “GIVE ME AN AGENT.”
I can’t help but think that whoever decided that automating the drive-thru experience was a good idea has either no idea how speech recognition technology “works” or has spent very little time ordering from drive-thrus. (Or both. Probably both.) It is exactly the perfect storm of bad speech recognition conditions: street noise, variable weather, faraway voices, multiple voices, accented voices. Let’s play out a scenario, shall we:
Drive-Thru Bot: “Hi, what can I get for you today?”
Customer: “I’ll get, uh … hmm. (to passengers in the back) Guys, pipe down; I’m trying to order”
Drive-Thru Bot: “Sorry, I didn’t get that. What can I get for you today?”
Customer: “What? No. I’ll get the, uh … Does the number seven come with mayo?”
Drive-Thru Bot: “A number seven with mayo! Would you like a drink with that?”
I can feel my blood pressure rising just thinking about it.
When I speak with Forrester’s clients about speech recognition technologies, the biggest hurdle is always accuracy. These systems tend to handle everyday words well enough, and the vendors may have done some tinkering to handle various languages, accents, and dialects. But when it comes to niche jargon and company-specific terms, you’re often working it up from nothing. Picture this: Deep in the heart of Boston, there’s a speech engineer training a speech model to interpret: “Ya, I’ll ordah a Woppah and a lahge coke.” And that’s just the tip of the iceberg.
Jokes aside, I wish companies spent more time considering the possible downstream impacts of these cost-saving measures. How many wrong orders will the in-store staff need to correct? How many times will an aggravated customer swing up to the pickup window to chew out an unsuspecting employee for their awful order bot? How often will a customer simply drive away, deciding that they’d rather eat somewhere else than fight to be understood by a machine?
These days, there really aren’t that many opportunities to connect with a customer one to one. In a world of automation, genuine human connection might be the secret ingredient that keeps customers coming back. Keep it real, folks!
Christina McAllister is senior analyst, Forrester Research, covering customer service and contact center technology, strategy, and operations.