TL;DR
I’ve come to rely on two acronyms to guide my AI usage around restaurants
BITE: Bring It To Everything
FAILS: F***king Around is Learning S**T
Some examples from Peppa Pig to Toast Sous Chef
SXSW Vote Reminder
Voting closes Sunday and I’d appreciate a vote if you haven’t (or maybe if you have?) and share. Thanks.
I am in the home stretch of planning the Industry Hackathon next week. The agenda calls for 45-60 minutes on the front end to talk through and demo concepts and practices around AI use that seem to work well. The pros at Agent.AI will do some of the heavy lifting here and I’ve pulled some other stuff from various newsletters and other resources I use a lot. My contribution to this is largely focused on two frameworks (maybe a bit of stretch to call them frameworks) that shape my approach. I figured I’d share them more broadly as well.
BITE: Bring It to Everything
It’s what it sounds like. Put a lot of stuff into AI to understand what it’s good at, what it’s not and, most importantly where it proves useful to you and where it doesn’t. Anecdotally it feels like most people go to one pole or another when they first use AI. What I mean by that is there are the people who use it once, get a crappy output and say ‘I don’t get what the hype is about, AI sucks’. The other pole is the people who use it, get a good output and then trust it for everything without question. Neither of these is a good idea since the right answer is somewhere in between. BITE helps you find your middle ground. Internalizing BITE has helped me test and keep up with the edges of AI capabilities.
Some of the most impactful, practical stuff we’ve done in the restaurants, like the Brunch line redesign, was 100% a result of BITE-ing.
FAILS: Fucking Around Is Learning Shit
I have a 3 year old (trying to watch my language) and she LOVES Peppa pig. Her grandparents wanted to buy her something so they asked me about the extended Peppa Pig family. I couldn’t remember much beyond Daddy Pig, Mommy Pig, Peppa, her little brother George and the new baby sister Evie. I asked ChatGPT for the whole family tree but it didn’t include Evie so I asked “What about Evie”. Here’s the conversation:
Not only did it not know about Evie. It told me I was wrong and then gave specific reasoning as to why I was likely wrong, going as far as blaming PP fan fiction.
This is illustrative of FAILS because I learned something important about how AI models work. They’re trained largely on internet data and the GPT-5 model I was using was clearly trained before the little miracle of Evie was born. Silly example to be sure but the bigger learning is actually important in understanding how and when to trust AI.
BITES & FAILS with Toast Sous Chef
A more practical use for me right now is in using Toast Sous Chef- which is their AI product. For a while I had to really BITE to stay with Sous Chef because it wasn’t very good. In the last 8 weeks it’s getting really, really interesting to use so I am fully in FAILS mode. Here’s a little back and forth in preparation for this Saturday night which we expect to be quite busy:
It wasn’t perfect but there were definitely helpful insights around earlier server call times, batching and to anticipate a late afternoon push. I don’t ever want our team to be so reliant on any technology that their own ability to anticipate and address challenges atrophies. If the tools are thoughtfully designed they will enhance our teams thinking, not erode it. For example, the Toast feedback triggered me to think about how we should hand the tricky line cook transition time at 3PM on Saturday. Can we see if anyone could stay and extra hour or come in a bit early to take smooth out the change over?
As you BITE and FAIL more, you’ll begin to know where to apply AI most efficiently and effectively. It’s important to feel for that and then check yourself before asking the AI to do something. Is this a good use for AI? Force yourself to ask that question to stay conscious both the environmental impacts of AI use and it’s potential to make you lazy (lazier) or dumb (dumber).
A vote for our SXSW panel is a vote for connection in a time of increased isolation.
I don't know how you find time to run restaurants, raise a family, explore AI, and also write about it but I'm glad you do