One of our goals with your client, Sanderson Farms, is to always look at ways we can try and help the consumer in honest ways. One of our main initiatives has been trying to help them find recipes that fit their needs and help reduce food waste with features like our “Search by Ingredient” tool for recipes.

At the end of 2023 I had the idea of trying to take further. To look into using one of the new large language model AIs to make a very natural language chat bot that would understand your questions and needs and help you cook your chicken. The team agreed, and after a few more internal cycles of refining the idea, Sandy was born.

We partnered with a development shop we had used a number of times in the past to look at taking our concept and building it out using OpenAIs APIs along with our own custom embedding and middleware allowing for a customized and pinpointed experience over the existing custom GPTs.

We tokenized our recipe database, cooking tips, and tricks. Then configured the bot to act as a virtual brand ambassador, assisting consumers with personalized recipe suggestions and cooking advice. We crafted custom starter prompts to give the bot our brand voice, and guard rails to lock it into our content and help understand the types of questions it would be fed. There is a custom middle layer that takes user’s message and breaks it down into actual queries to help ensure it only provides accurate results from our site.

After multiple weeks of trial and error, refining various prompts/context/embeddings, and really trying to get the bot to break character we ended up on a solution everyone was happy with. Overall, the bot enhances customer engagement by providing instant, interactive support, making it easier for users to explore Sanderson Farms' products while delivering a storytelling experience that strengthens the brand-consumer relationship through seamless and intelligent interactions.