There’s been an explosion of ideas on how advanced artificial intelligence can improve restaurant operations in recent years. One area that’s generated its fair share of attention is voice AI in the drive-thru, which uses machine learning to process, interpret, and respond to spoken language.
Many quick-service brands started experimenting with the technology after the pandemic. Now, many of them are scaling their systems beyond pilot programs and into widespread deployment.
Bojangles is one of those chains accelerating its investment in voice AI at the drive-thru. Following a successful pilot with tech provider Hi Auto, the company this summer inked a deal to bring the technology to hundreds of stores across the country. Its AI ordering assistant already is in over 200 locations, with new ones being added almost daily.
MORE: Check out the full 2024 QSR Drive-Thru Report
The system, called Bo-Linda, uses natural language processing and voice recognition to manage orders. It also gathers information from employees, like wait times or out-of-stock alerts, and relays relevant updates to customers.
“I think a lot of people went into this the same way that we did, thinking about it in terms of a cost offset device or as a way to deal with a shrinking labor force,” says chief information officer Richard Del Valle. “It absolutely fulfills on those fronts, but what we’ve discovered throughout this whole journey is that there are other benefits that are even more important to us.”
As an example, he points to Bo-Linda’s ability to reduce guest anxiety and confusion by confirming that customers are promptly greeted, eliminating concerns about whether their arrival has been noticed or when they should begin placing their order.
“Think about everything you want to accomplish with the order-taking step,” Del Valle says. “You have to engage the guests in a meaningful way. You have to be able to answer any questions they might have. And you have to make sure they get exactly what they want.”
Factor in fast food’s notoriously high turnover, and that leaves most restaurants in constant training mode to certify those boxes are always checked.
“There’s also an art to talking to guests with real warmth, to upselling and encouraging them to take advantage of offers and promotions,” Del Valle says. “Getting that every time from the workforce is a real challenge. This takes all of that out of the mix.”
Several factors can lead to poor customer satisfaction in the drive-thru, whether it’s inaccurate orders, long wait times, or a lack of hospitality at the window. Taco Bell has seen first-hand just how effective voice AI can be at avoiding the stressors that lead to those problems in the first place.
The brand started piloting automated ordering in 2021 with a handful of locations in its home state of California. The technology is now supported by digital menu boards as well as parent company Yum! Brand’s proprietary POS system. Soon, it will be integrated with the chain’s loyalty program, too.
The brand’s voice AI system currently is up and running at over 100 stores, including several units owned by Diversified Restaurant Group, one of Taco Bell’s largest franchise partners.
“When I go in, I try to confuse the AI and order stuff that no one else would think of—putting chicken on a breakfast item or adding with pickles,” says Todd Kelly, COO of Diversified Restaurant Group. “They get a little bit confused with the voice AI, but they’re always friendly.”
The biggest benefit has been assuring stores deliver the same top-notch experience every time a customer cruises through the channel, he adds.
“The voice AI doesn’t have a bad day,” Kelly says. “They’re not in a bad mood when they come to work. They don’t get hungry. They don’t need a cigarette break or whatever. They keep plugging away. So, it’s nice from that standpoint. And they never miss trying to round up to our donations, a big part of who we are in the communities that we serve. They never forget to offer a drink to our customer. There are some bumps, but we’re actually expanding our test in all of our markets. Within the next couple of months, we’re going to have voice AI in about 30 of our restaurants.”
Taco Bell indicated this summer that it plans to expand the technology to hundreds of more locations over the same period. In a statement announcing the accelerated rollout, Yum! Brands’ chief innovation officer Lawrence Kim said he’s confident in the tool’s ability to deliver meaningful benefits around customer satisfaction and improved operations, citing over two years of fine-tuning and testing out on the West Coast.
Wendy’s is also making strides with its drive-thru AI initiative, dubbed Wendy’s FreshAI, which launched last year through an expanded partnership with Google Cloud. It created the platform to overcome the challenges of traditional automated systems, which often struggle with complex customer interactions, uncommon phrasing, and complicated requests.
Unlike those older rule-based systems, Wendy’s says its technology uses generative AI to formulate dynamic responses and adapt in real time.
Since December 2023, Wendy’s FreshAI has expanded from four company-operated restaurants in the chain’s home state of Ohio to 28 restaurants across two states, with further expansion on the horizon. Results have so far been promising. An early pilot in Columbus, Ohio, showed a 22-second improvement in service times compared to the average for that market.
In August, the company introduced Spanish language capabilities to further enhance customer engagement as it continues scaling the technology across its footprint.
“I’m energized by our partnership with Google Cloud to continue pushing this technology forward,” says chief information officer Matt Spessard, adding that Wendy’s will continue exploring new ways to expand access and reshape the drive-thru experience going forward.
Beyond improving service speed and accessibility, Wendy’s is placing equal importance on how the system impacts team members. The company’s technology and operations team have clocked thousands of hours working from dining rooms and engaging with crew members to ensure the system improves the experience for store-level workers as well as the guests they’re serving.
For Bojangles, the positive response from team members working alongside Bo-Linda is another key advantage. Del Valle says it far outweighs the cost-savings that could come from using the tool to operate drive-thrus with less staff.
“You’re taking roughly 35 percent of the workload away from one or two employees by removing the order taking, which then frees them up to be better at speed, better at accuracy, and better at cheerfulness at the window when they engage with guests to take the payment,” he says. “We’re finding that our employees are very happy with that, to the point where if you walked into one of our Bo-Linda restaurants and announced to the crew that you’re taking it out tomorrow, they’d be very upset. They’ve grown to like it and count on it. It’s fun to see just how quickly this technology becomes part of the team.”
White Castle also is pushing back on the widespread notion that voice AI in the drive-thru is meant to replace human jobs. In fact, the idea that it could free up staff members to focus on engaging with customers is what first sparked the company’s interest in the technology four years ago.
“It hasn’t affected our employment levels one bit,” says Jamie Richardson, White Castle’s VP of marketing and public relations. “If anything, it’s given us more capability to invest in our people. It’s not about the pathway to greater productivity. It’s not about cutting hours. It’s about giving our people better tools, so they’re able to take better care of customers, and then all things being equal, you’ll get more return visits sooner, and that math just keeps playing out over time.”
The sliders chain initially tested the technology in 2020 through a partnership with Mastercard, with SoundHound providing the voice recognition tool and another company offering vehicle recognition software. Now, it is working directly with SoundHound on an AI assistant it calls Julia, named after a beloved White Castle hostess from the 1930s.
The company has been steadily expanding Julia into more stores and plans to have it installed in 100 locations by the end of the year. Richardson says there have been plenty of learning moments along the way, especially when it comes to ensuring the system recognizes the vast amount of words and expressions people can use to communicate one simple message.
“We had no idea how many different ways customers can choose to order the same thing, like ‘give me some of this,’ or ‘I’ll take that,’ or ‘how about this?’” he says.
The potential variations multiply exponentially when you start dealing with customizations, clarifications, and revisions. Richardson says dialing into those nuances was daunting at first. It was a bigger challenge than the company initially anticipated. But it also ended up being a key proof point for continuing to invest in the tool.
“It was amazing to see how quickly Julia could adapt and take on all of these new learnings,” he says. “Once we were able to recalibrate and say, ‘alright, it’s going to take a minute longer than we thought to get all the different nuances down’ it was just so outstanding to us, because Julia only gets faster and faster at learning.”
Not all fast-food chains have found success with AI in the drive-thru. Notably, McDonald’s this summer decided to end its trial of IBM’s Automated Order Taker due to difficulties with interpreting accents and dialects, which compromised order accuracy. At the time, the company said it would explore other vendors with the goal of reaching a decision on a future voice-ordering solution by the end of 2024.
The fast-food giant reportedly couldn’t get its non-interaction success rate—the percentage of how many orders are handled by AI without any restaurant team member intervention—out of the low 80s.
“If I were in the low 80s, I’d shut it off too,” Del Valle says. “It’s counterproductive to have your team constantly stepping in.”
Bojangles is achieving around a 97 percent non-interaction success rate, a figure he attributes to extensive work conducted during the setup phase.
“The trick is that whatever comes out of a guest’s mouth, the system has to be able to fulfill,” he says. “I think that’s where a lot of people trip up and where you tend to see the most failure.”
Another challenge is determining the optimal time an AI voice system should pause before prompting customers. Restaurants need to ensure it’s long enough to be polite and accommodating but not so long that it slows down the service and frustrates guests.
Richardson says it also needs to strike the right balance between speed and efficiency while maintaining a touch of personality.
“We’ve got a little bit of that right now with different phrases that are less generic and recognize you’re at a White Castle,” he says. “One of the other big things we found early on was that the speed of Julia’s voice was too slow. We needed to speed it up to be a bit more conversational and make it sound less like a machine. It was really interesting to see how much better that made it. It’s the little things like that that end up making a big difference.”
Looking ahead, he views voice AI as more than just a flash-in-the-pan trend. He sees it as a significant development with the potential to transform how the quick-service industry handles its leading sales channel.
“There was a time when there was no such thing as self-service at the gas station or self-checkout at the grocery store,” Richardson says. “Then technology improved and provided more options. The same thing is happening here for restaurants. If we’re all doing our jobs in our industry, we’re going to ask good questions, listen intently to what our customers tell us, and then be responsive to those needs. It really feels like we’re on the right track in this instance, so I think it’s going to become the norm rather than the exception.”
