How Starbucks could boost profits by putting AI chatbots at its drive-thru windows

A lot has been made about the benefits of artificial intelligence, particularly as it relates to how big cloud companies such as Amazon (AMZN), Alphabet ‘s Google (GOOGL) and Microsoft (MSFT) are developing the software, and how hardware companies like Nvidia (NVDA) are making the brains that make all that AI software possible. That’s certainly been a boon to all of those Club names as their stocks have soared this year. However, what hasn’t got as much attention is the opportunity that generative AI applications can bring to other sectors in the form of increased productivity and efficiency — and ultimately, margin improvements. The quick-serve restaurant (QSR) industry, for example, is an area poised to see huge tailwinds from AI. Wendy’s (WEN) and Jack in the Box (JACK) certainly think so as vocal, recent early believers. For the Club portfolio, our QSR name is Starbucks (SBUX). We think the coffee giant could increase profitability through the use of generative AI at its drive-thru windows, which would allow the chain to reassign difficult-to-find workers within their current locations to tasks that require the human touch and/or to redeploy them to open up new stores. Why the QSR industry? For starters, we’ve already seen what Wendy’s wants to do. In a new partnership with Alphabet, the burger chain plans to use a custom AI chatbot at drive-thrus. CEO Todd Penegor told Jim Cramer in a May interview for “Mad Money” that he wants to “lean into late night,” which has typically been an underserved daypart. Wendy’s will pilot its new AI drive-thru tech this month at some locations in the Columbus, Ohio, area. Jack in the Box has also been looking into ways that artificial intelligence can make the company more efficient at a time when inflationary pressures on the economy have increased food costs. CEO Darin Harris told Jim recently that his team is looking to enhance profitability via an improved labor model that leverages robotics and AI. In fact, Harris said that a voice AI solution currently used at company Del Taco stores is already taking about 85% of the orders at the locations where it’s been deployed. The AI is proving more effective than its human counterparts at upselling — convincing customers to buy additional or more expensive menu items. In an attempt to extrapolate how such AI applications could help Starbucks, we first wanted to recognize the drag that the ongoing labor shortage in the U.S. has had on hiring, which has left many food service businesses scrambling to staff the locations they already have — let alone expand. Adopting a Wendy’s-like AI model at Starbucks’ drive-thrus could alleviate some of the hiring issues by shifting current would-be drive-thru workers into different open roles or ones that can’t be done by machines. The AIs could even expand accessibility and grow the customer base by recognizing many different languages (besides just English), including sign language. Drive-thru locations, as we see it, make perfect sense as a starting point when considering the relatively controlled nature of the experience: a limited, largely set menu with orders generally placed by speaking to a person over an intercom, not face-to-face. Building new drive-thru locations with AI in mind from the ground up could increase the rate of expansion as it would require less training and a reduced need to find new labor. That’s because, traditionally, a new drive-thru would require new employees working in multiple dayparts to take customer orders. However, with generative AI, a one-time upfront cost — along with ongoing maintenance charges far below the cost of a human worker’s salary and benefits — could significantly increase the profit margin on those new locations. Sure, there may be some deviation from time to time when people order. But, given enough encounters, all of which can be used to train and customize large language models used in such AI tasks, it stands to reason that the vast majority of interactions can be handled by chatbots. While citing the “human experience” as a reason AI won’t disrupt face-to-face service roles, some of the critics of automation may not be fully appreciating how convincing these large language models are fast becoming. We have already seen a demonstration of this technology over a year ago when Nvidia unveiled ” Project Tokkio .” However, all these months later, it’s only really now becoming more clear just how capable this solution can be with the more widespread availability of generative AI such as OpenAI’s ChatGPT, which launched and went viral late last year. OpenAI is backed by Microsoft. Another recent entrant is Bard, Alphabet’s answer to ChatGPT. Ordering kiosks in QSRs are nothing new, but they’ve generally revolved around a touchscreen interface. However, with the introduction of generative AI and tools that allow companies to train large language models for their unique business needs, the line between a human experience and ordering from a digital touchscreen kiosk — similar to placing an order online — is increasingly blurred. That ability to maintain something of a human interaction with the convenience of digital should help companies increase their digitization efforts without sacrificing the customer experience. That’s a huge factor in the QSR industry, given the number of options available to consumers. It’s also a much better solution in certain instances, such as drive-thrus, where a touch screen may be an inconvenience. In researching this thought experiment, we did reach out to Starbucks for comment. In regard to the implementation of artificial intelligence, management appears to be largely focused on its Deep Brew initiative laid out a few years back. The idea is to leverage AI not as a replacement but as a “super-smart sidekick to the humans wearing the green aprons,” Starbucks said, with a focus on inventory management, supply chain logistics, and drink replenishment to name a few uses. Though we did ask about any newer initiatives revolving specifically around generative AI, mentioning the experimentation going on at Wendy’s and Jack in the Box, the company said its core focus right now is Deep Brew. The hesitancy to discuss any generative AI implementation is understandable. After all, what made Starbucks the incredible company that it is today is its desire to remain connected to its employees and maintain the human element of the Starbucks experience for customers. That said, we think that generative AI opportunities at Starbucks are still worth considering. At the end of the day, Starbucks is a public company with shareholders, the Club among them. And, therefore it must stay competitive. Implementing generative AI doesn’t have to mean layoffs, it could simply mean increased operating leverage and a greater opportunity for expansion both in the U.S. and abroad. It’s reasonable to believe that should the competition have success with their own generative AI initiatives that management will have no choice but to consider it themselves or risk being disrupted. Thought experiment At a high level, it’s pretty easy to see the benefits that generative AI can bring to QSRs such as Wendy’s and Jack in the Box — or even, Starbucks. However, to get a better sense of the financial opportunity, let’s look at some numbers. According to Starbucks’ fiscal 2022 annual report, wages and benefits accounted for $8.16 billion. Company-operated locations generated $26.56 billion (which was about 82% of total fiscal 2022 sales.) Doing the math, the operating margin at company locations was 17.4%. So, we can see that as far as company-operated stores go, Starbucks spends just over 30% of sales on wages and benefits. This is where we think there’s a clear opportunity for margin improvement as generative AI adoption goes mainstream. You can’t cut out the entirety of that $8.16 billion expense. You still need humans to prep the food and drinks, clean, answer questions that a large language model may not be equipped for, and so on — basically do all the jobs AI can’t replace. Though longer-term, these jobs too may be at risk of disruption from AI, given that we have already seen several drink-mixing robot solutions. However, we can start to get a sense of just how impactful even a small amount of adoption can be to the bottom line. Looking at the Starbucks table we put together, you’ll see an estimated breakdown of positions for an average location and wages and benefits BEFORE and our estimates for AFTER in a generative AI scenario. If Starbucks were to replace 19% of its in-store U.S. workforce — three baristas and one shift manager, as represented by the difference Per Location before AI and Per Location w/AI column — the company could decrease its U.S. wages and benefits expenses by a little over $1.06 billion. Apply that same logic to International — represented by China’s Per Location before AI and Per Location w/AI estimates — and that’s another $193.33 million for Global Total savings of $1.26 billion, as seen in the green box in the bottom right corner. (The wage and benefit numbers were estimated using figures from Glassdoor.) Put another way, there are, on average, about 21 employees per company-operated location, bringing that number down to about 17 — either through attrition or redeployment to open more stores — would potentially expand the company-run store operating margin by about 4.7 percentage points to 22.1%. That may not sound like a lot. But, given Starbucks generated operating income of about $4.62 billion in fiscal 2022, we are talking about a roughly 27% increase in the total operating profit. This is admittedly some back-of-the-envelope math, with headcount sourced from company filings and salaries sourced from Glassdoor and adjusted to normalize for part-time workers. The employee makeup was also estimated based on feedback sourced from the Starbucks subreddit. However, we think it’s sufficient to demonstrate the profit opportunity open to companies such as Starbucks that have high headcounts in roles that generative AI solutions can and already are seeking to address. The adoption of generative AI for select roles removes the costs associated with hiring and training new employees and serves to increase operating leverage as the cost to train a large language model customized for the Starbucks experience shouldn’t see much cost increase as new locations are opened. It also increases accessibility thanks to the ability to serve customers regardless of what language they speak. While it remains to be seen what roles can indeed be replaced by AI, without risking any damage to the Starbucks experience, there remain many unanswered questions in these early days: will people take to ordering from machines, how convincing can an AI be, will hardware robotics prove reliable for those tasks that require a physical component, or what new issues will arise from the use of AI? We think this thought experiment to be at the very least directionally accurate. Bottom line Acknowledging that AI can replace certain roles is not necessarily a call for layoffs. Rather it’s a call for increased operating leverage. Though we are currently focused on the QSR industry and more specifically on Starbucks, what we want members to take away is that there’s a massive opportunity for margin expansion across industries outside of the technology sector. We think that companies with a large number of employees in roles that AI is already demonstrating an ability to disrupt are the low-hanging fruit for margin expansion opportunities. Companies that rely heavily on manufacturing are also likely going to realize the benefits further down the line than those that can use a software-only solution. There are going to be a lot of opportunities to invest in to play the adoption of generative AI outside of tech. As we scan the market for opportunities we will do so with this AI tailwind in mind. While the market is intently focused on the creators and the enablers of AI, investors may be missing out on the benefits to the customers these developers will ultimately service. The profit opportunity outside of tech could prove even larger than the one in tech right now. (Jim Cramer’s Charitable Trust is long AMZN, GOOGL, MSFT, SBUX, NVDA. See here for a full list of the stocks.) As a subscriber to the CNBC Investing Club with Jim Cramer, you will receive a trade alert before Jim makes a trade. Jim waits 45 minutes after sending a trade alert before buying or selling a stock in his charitable trust’s portfolio. If Jim has talked about a stock on CNBC TV, he waits 72 hours after issuing the trade alert before executing the trade. THE ABOVE INVESTING CLUB INFORMATION IS SUBJECT TO OUR TERMS AND CONDITIONS AND PRIVACY POLICY , TOGETHER WITH OUR DISCLAIMER . NO FIDUCIARY OBLIGATION OR DUTY EXISTS, OR IS CREATED, BY VIRTUE OF YOUR RECEIPT OF ANY INFORMATION PROVIDED IN CONNECTION WITH THE INVESTING CLUB. NO SPECIFIC OUTCOME OR PROFIT IS GUARANTEED.