The increasing integration of AI into daily life has sparked global discussions, particularly around the topic of job displacement due to AI advancements. Interestingly, in some sectors, AI's capabilities have been overestimated, leading to its premature retirement.
Recently, McDonald's announced the termination of its AI ordering project. This automated ordering system, developed in collaboration with IBM, was deemed unsuccessful after a three-year trial. Beginning in 2021, McDonald's and IBM tested an AI-based automated voice ordering system in over 100 McDonald's drive-thru locations across the United States. Traditionally, drive-thru customers order at the first window, pay at the second, and collect their food at the third, then drive away. With AI integration, the drive-thru process aimed to streamline this into a single step.
Despite the convenience promised by AI ordering, McDonald's has decided to discontinue it due to its poor performance. According to reports from U.S. media, the AI ordering technology provided by IBM is still in its nascent stage, with a voice recognition accuracy rate of only 85%. This means one in five orders required human assistance. Unfortunately for McDonald's, despite IBM's technology being suboptimal, the fast-food giant had little choice but to proceed, as the AI ordering collaboration was a byproduct of another acquisition deal between the two companies.
As part of its “digital transformation,” McDonald’s acquired Apprente in 2019, a company focused on developing technologies for complex, multilingual, multi-accent, and multi-task conversational ordering. McDonald's then established McD Tech Labs based on Apprente. However, McD Tech Labs failed to meet McDonald's expectations, leading to the 2021 agreement where IBM's acquisition of McD Tech Labs was a prerequisite for their AI ordering partnership.
In a way, the collaboration between IBM and McDonald’s can be aptly described as “throwing a feast for a spoonful of vinegar.” McDonald’s, as a traditional restaurant chain, struggled with AI, rendering the expensive McD Tech Labs investment unproductive. The McDonald's leadership at the time might have thought that if they couldn’t make it work, selling McD Tech Labs to IBM and letting IBM provide the tech support was a better option. Unfortunately, what McDonald’s couldn’t achieve, IBM also failed to accomplish.
In truth, AI's attempt to revolutionize the restaurant industry has largely become a graveyard for tech giants, with numerous big players investing heavily only to yield no tangible outcomes. The core of AI-powered ordering is cost reduction and efficiency enhancement, but in practice, it often leads to dual declines in these areas. The advent of advanced language models like ChatGPT late in 2022 marked a new era in natural language understanding, yet even with these advancements, AI ordering systems struggle.
While large language models (LLMs) were nascent during McDonald's AI project, ChatGPT differentiated itself from traditional voice assistants like Siri with its ability to engage in extended conversations, demonstrating a more nuanced understanding of context and emotion. However, in the fast-paced environment of quick-service restaurants like McDonald's drive-thrus, where speed is paramount, the lack of sophistication in these AI systems increases customer interaction costs, making them less efficient than scanning a QR code to order.
The primary pain point of AI ordering in practical use is not the AI's understanding ability. With the support of large language models, AI's ability to comprehend human language has fundamentally improved compared to two years ago. The real challenge lies in the far-field voice recognition technology, which struggles to accurately capture target user voices from complex acoustic environments. Background noise and reverberation can drown out users' voices, resulting in AI systems being unable to hear users clearly, necessitating frequent human intervention.
In conclusion, despite continuous advancements in AI technology, its practical application in the restaurant industry, particularly in fast food scenarios that prioritize speed and efficiency, has not met expectations due to technical immaturity and operational challenges. This underscores the need for the industry to more carefully assess the maturity of technologies and their suitability for specific applications when advancing automation initiatives.
