Wednesday - last edited Wednesday
My name is Eric, and I’m the owner of Old Scratch Pizza in Dayton, OH, and a Toast Customer Advisory Board member. AI has become a daily tool in our restaurant, helping with everything from employee communication and marketing to streamlining operations and reputation management. Tools like ChatGPT and Claude assist with writing, problem-solving, and automating tasks, saving time and effort.
I also use AI to extract deeper insights from Toast data, giving us answers to questions that were previously unavailable. If you’re curious about AI in restaurant management or reporting, I’m happy to share my experience—ask me anything!
Wednesday
Hi Eric,
Thanks for taking the time to answer some questions! What level of trust do you put in the validity of AI analysis with programs like ChatGPT? As a followup, what do you do to double check the work of the AI to insure you are making decisions based on correct analysis rather than what from time to time is incorrect information from AI?
Best,
Brian
yesterday
Thanks for the question! I think about validation in two areas: data accuracy and factual correctness.
For data—typically spreadsheets and formulas—errors usually stem from how I structured my prompt. The great thing about AI is that it’s a conversation. If the output doesn’t seem right, I ask the system to check its own work or explain its reasoning. If I still see an issue, I refine the prompt and try again. One trick I use is asking, “What prompt can I use to get this result every time?”—this helps create reliable, repeatable queries. We’ve built prompts this way for weekly reporting, like analyzing fulfillment times in Toast.
For factual accuracy—especially research—I like using Perplexity.ai. It provides well-researched answers with direct source links, making it easier to verify information. AI is a great tool, but knowing how to validate its output ensures you’re making informed decisions. Hope that helps!