Back in February, Yum! Brands, owner of the likes of KFC, Taco Bell, Pizza Hut, and Habit Burger & Grill, launched its own AI-driven restaurant technology platform, Byte By Yum! As per the pitch at the time of the announcement:
The implementation of Byte by Yum! will also enable a faster and more impactful adoption of AI by Yum! and its brands. Backed by Artificial Intelligence, Byte by Yum! offers franchisees leading technology capabilities with advantaged economics made possible by the scale of Yum! The Byte by Yum! platform includes online and mobile app ordering, point of sale, kitchen and delivery optimization, menu management, inventory and labor management, and team member tools. With the introduction of Byte by Yum!, the company is phasing out the legacy names and branding for technology products such as Poseidon, Yum! Commerce Platform, Tracks, and SuperApp, which are now housed under Byte by Yum!. In the US, all four of Yum!'s brands are using elements of Byte by Yum! to process more than 300 million digital transactions a year.
It's an innovation that Yum! Brands sees as perpetuating digital leadership in the highly-competitive Quick Service Restaurant (QSR) space. Joe Park, Chief Digital & Technology Officer at the group, notes:
Over the last five years, the world's accelerated into digital . The restaurant industry is no different. Back in 2019 we were at about 19% digital sales as a percentage of total sales; if you look at year-to-date, we're at over 50%. So it's massive acceleration...We're probably more of an e-commerce company than a fast food restaurant company in many ways.
An important decision for Yum! Brand was that it had to be an AI-first operation, he explains, noting that pre-COVID there was effectively a duopoly when it came to delivery of takeaway food that basically boiled down to Chinese or pizza. That's all changed:
What's happened is now we've got third party aggregators who make it so you can get Thai food, you can get sushi, you can get all kinds of food, and it's made it incredibly competitive. Think about all the data now that's required. Think about all the channels we've now opened up and the complexity to run a restaurant. Clearly, we're going to rely on data, and, even more obvious, you're going to have to make sure that AI can help us make the best decisions we can drive operations.
For most restaurant brands, technology is now critical and complicated, he argues:
The reality is, if you're a restaurant general manager, you probably have at least 15 different technology vendors. It's like an alphabet soup of, you need a CMS, ERP, you need to have a CDP, you need to have all these different [things] and the odds of us being able to integrate them all and thinking they'll be there for the long run are really slim. In this AI-first world, it became clear to us in our vision that if we didn't offer a 'restaurant-in-a-box, where we had a Point-of-Sale, an e-commerce platform, back-up technology, mobile apps that not only integrated altogether, but had a UX that was based on human-centered design so that a team member coming in could use them intuitively, it was just gonna be a losing bell.
Which is where Byte by Yum! comes in:
It is really about having this integrated platform that makes it super-easy to run a restaurant and serve our customers, and ultimately powered by our own internal, proprietary AI platform that we call Crave AI...For us, it starts with the belief that we can engrain AI into all parts of a customer's journey with one of our four iconic brands, and also for our team members in terms of how they go about their jobs. If you start with that belief and try to solve the business challenges, one of the ones that we often take a look at is to say, 'Could we help our team members jobs easier? Can we help improve the speed of service? Can we help improve order accuracy?'
Park cites the example of voice AI:
Between all of our brands, we've probably tested voice AI in many forms over the past several years. Although we love working with third parties, and we've had numerous success piloting them and even scaling some solutions...[we feel] there's got to be a better way. How do we move away from this black box of training AI models to get voice AI to be faster, more accurate and more human?, How do we think about scaling this so that it makes sense for us and supportable store-by-store-by-store, especially as you go up [in scale].
Pizza Hut is a good case in point:
If you call your local Pizza Hut and they're really busy, it'll overflow to a call center. We believe that there's a really powerful use case there if we could apply voice AI to the call center...We do think that can be a better experience for customers because voice AI is always positive, has the right tone of voice, has the right word selection, can up-sell consistently as we so choose. We think that the latency has gotten so much better with technology as it goes through cycles.
Yum! Brands is partnering with NVIDIA to drive forward innovation here, he says. Another useful example around voice AI comes from Taco Bell and its use of "foreign, made-up words" on its menu, terminology that you're not going to find in Webster's English Dictionary:
The odds of voice AI, especially when we think about ASR (Automatic Speech Recognition), being able to pick up on it, that's hard. [We are] partnering with the NVIDIA product team on getting the lexicon right, coming up with the custom dictionary, being able to fine tune it so that we can account for different ways the diverse folks of the world pronounce it. It's so important getting the getting past the uncanny valley of the tone of voice AI. It'll also be really important to make sure it sounds like a human, to get past the unease folks might have. [With] pronunciation, when it comes to phonemes and being able to get every single syllable right, all that adds up.
Coming up, of course, will be greater use of agentic AI and Park sees this as a supporting and enabling technology for restaurants human employees by simplifying cognitive overload:
You're trying to listen to the customer over a speaker that might be at eight kilohertz, really low-fi ,and you're trying to make sense of what they say at the same time, punching in what they want at the same time for an up-sell. That's really hard to do - 'Hey, welcome to Taco Bell. Could I please take your order?'. Your brain kind of shuts off and goes into cruise control, and you put 100% of your brain power into getting the rest of the job right...Having the agent in the workflow, it's making a huge difference in getting our team members to feel that these are easier jobs than before.
All of this innovation has come with a price tag, of course, but it's clear Park sees it as money well-spent:
Since we started our digital technology journey over six years ago, as a company we've invested over a billion dollars into building out and ramping up our capabilities. Today, we've got close to about 2,000 digital and technology employees within Yum! Brands. When you start going through these numbers, you go, 'Hey, that's that's not a typical QSR company. When you think about the scale of the talent, the size of the investments, and some of the capabilities we've either built internally or acquired in the form of start-ups.
Our approach is that it's important to have this talent density, making sure we've got the best technologists who we attract externally or who we build up internally. I get very proud and excited about the leadership team we've built out in order to drive this vision forward. For us, this whole philosophy is that we believe that by owning our destiny in technology, so to speak, we can likely deliver it better, faster and cheaper than any other way for our franchisees across the world.
And an important learning that's come out of all the transformation activity is that there is a need to 'budget 'for humanity in the process, he suggests. Again, voice AI pilots provide a good example in practice as some stores report 80% accuracy rates, others much less. In the latter category, the danger then is that the team in-store by-passes the AI in order to get on with their jobs, rather than persevering and giving the tech time to learn and improve.
If every store manager had that mindset of letting things play out, the number of AI use cases would accelerate, ponders Park:
