Wendy’s Serves Up Generative AI To Boost Its Customer Experience

How generative AI changes the quick service restaurant industry

Every industry, including the quick service restaurant (QSR) market, plans to transform its business with artificial intelligence (AI), especially generative AI. Several years ago, Wendy's embarked on its AI journey, leveraging cloud computing services and generative AI to enhance employee and customer experiences. The drive-thru experience presents numerous challenges for QSR restaurants due to the complexities of menu options, limited-time offers, special requests, and ambient noise.

Wendy's chose to tackle the drive-thru experience with AI because 75 to 80 percent of Wendy's customers choose the drive-thru as their preferred ordering channel. The company saw a tremendous opportunity to improve the customer experience by creating a seamless ordering experience using AI automation in the drive-thru.

In an interview with Maribel Lopez of Lopez Research, Wendy's CIO Matt Spessard shared how its AI program had advanced over the past year and shared advice for other leaders looking to tackle AI within their business. In 2024, Wendy's announced an expansion of its partnership with Google Cloud, tapping into Google's AI technology and resources to enhance its generative AI platform called Wendy's FreshAI. Wendy's Fresh AI aims to address challenges that traditional AI couldn't solve, such as understanding casual conversations and handling the extensive customizations of Wendy's menu. For example, the back-and-forth nature of conversations is an intensely complicated technical challenge for AI. It also added Spanish as a language option in 2024.

Traditional rule-based AI chatbots weren't the answer because they can't easily support natural conversations' diverse and dynamic nature. For example, Wendy's realized there are over 200 billion combinations of words and options to order a Dave's Double. Additionally, it can take years of development and tremendous work to maintain, modify, and expand capabilities within these more rigid rules-based solutions. Today, Wendy's uses generative AI to interpret conversations, create responses, and adapt in real-time instead of following a narrow set of rules.

While many AI discussions focus on job loss issues, Wendy's shared that its AI efforts also benefit its employees. Wendy's FreshAI works alongside restaurant teams, eliminating ordering issues while empowering crew members to focus on preparing and completing orders efficiently.

How does Wendy's measure AI success and return?

Wendy's takes a pragmatic view of what AI success means. Its efforts focus on delivering speed, accuracy, and consistency to customers. It measures these results with metrics such as how many orders were submitted without human intervention and the consistency of the customer experience at the drive-thru. What have been the results thus far? The percentage of orders successfully handled by Wendy's FreshAI without restaurant team member intervention averaged 86%, and it expects the average to increase.

One test site showed service times 22 seconds faster than the Columbus, Ohio market average. Wendy's noted that other QSR companies define "accuracy" as any order started by the AI assistant and submitted to the point-of-sale system, including orders where a crew member joins the conversation to correct an inaccuracy. Wendy's shared that if it uses that broader definition of accuracy, its FreshAI success rate reaches nearly 99%.

What's next for AI at Wendy's

Wendy's FreshAI has moved from pilot to production. It is now available in nearly 100 restaurants across 17 states with Spanish language capabilities. There's still tremendous upside to expanding the technology deployment as Wendy's operates over 7,000 restaurants globally. Going forward, Wendy's sees the Fresh AI assistant as a platform that will scale across various ordering channels, such as mobile apps, kiosks, and smart devices.

Advice on tackling AI innovation?

Wendy's CIO Spessard said the company learned many lessons along the way. He provided the following key pieces of advice for organizations looking to adopt AI:

1. Identify the right use case and technology. Wendy's emphasizes that organizations need to think pragmatically about the use cases where AI can provide the most value and then select the appropriate AI technology (e.g., generative AI, traditional machine learning) to fit that specific need. AI is a broad spectrum of technologies, so the "right tool for the right job" approach is essential.
2. Start with small-scale experiments. Spessard recommends that companies start with small-scale experiments within the organization, exposing employees to AI capabilities. This exposure helps organically grow the usage of AI as more use cases are identified. Spessard said Wendy's had to resist the urge to deploy too widely and quickly. Iterating in the early stages of its AI deployment helped Wendy's achieve its desired accuracy and consistency.
3. Emphasize continuous improvement. Wendy's stressed the importance of a feedback loop, continuous iteration, and a mindset of even 1% improvement every day. It regularly analyzes customer and employee feedback, allowing Wendy's to refine aspects like tone, cadence, and phrasing in Wendy's FreshAI. These refinements were critical to enhancing the AI assistant's performance.
4. Engage stakeholders early and often. Wendy's found it very valuable to regularly perform demonstrations of the AI assistant's capabilities and progress to key stakeholders like franchisees. These demos helped build trust and support for the initiative as stakeholders could see the improvements over time. Additionally, Wendy's noted that the rapid democratization of AI technologies is making change management easier, as employees are already familiar with using AI in their personal lives, setting the expectation for it in the workplace as well.

A key takeaway from the discussion was to ground AI innovation in iterative improvement, continuously engage employees and partners in the process, and cultivate a willingness to experiment and learn rather than trying to perfect the technology before deployment.