Wendy's use case of a customized language model is one that other fast food and fast casual restaurants might pick up. Placing the order is often the longest part of the process and not the most enjoyable one. If a chat-bot could engage the customer better, be friendlier and chattier and help them make their choices more effectively, that could help outcomes for both sides. Those preparing the order would have confidence that the order accurately reflects what the customer asked for. The customer could experience less friction in the flow. I was at a Cava recently pretty late in the evening. They were Missing some items for my order so there was back and forth on substitutions. 

The same thing happened with three other customers in line behind me. With a Chatbot, it would not be so hard to present the customers with options that the currently available ingredients will support. The bot can also recommend the right substitutions for your favorite item in the menu when some components are missing. All this can happen without stressing out the employees who have to prep the bowl while dealing with a hungry customer, missing ingredients and substitutions. In this kind of collaboration between bot and employee, it seems everyone comes ahead including the customer. Back to what Wendy' - not sure if the solution calls for custom LLMs, seems like an overkill for the job at hand.