BTW, the image looks nothing like Peter, except for the chill stride.

I handed the AI budtender to an actual budtender. He called it basic, told me not to make it more complex, and gave me the best design feedback I've had so far.

By April the budtender app worked. It answered questions, recommended products, and had a UI I was reasonably proud of. What I didn't have was anyone who does the actual job looking at it. So I sat down with Peter, a former budtender at our shop who was considered among the most knowledgeable on the floor.

Peter's initial reaction: "Basic"

His overall verdict was that the app is basic, in a good way. It functioned. I took it as a fair description of a first version.

Then he said two things I wrote down word for word.

"Our customers are potheads. They don't wanna read. They just wanna poke at a tablet."

And, in the same breath: don't make it more complex.

I came in wanting a feature list. I walked out with a constraint, and the constraint is more valuable. Chips over typing. Short over long. Pictures over paragraphs. I had already been leaning that way, but there's a difference between a hunch and hearing it from the person who spent his days with customers.

Focus on the middle half

Another budtender had given me a rough split earlier: about a quarter of customers know exactly what they want, about half are open to suggestion, and about a quarter are first-timers who are a little anxious. Peter landed on the same numbers without being prompted.

He also told me the middle half is where he earned his keep. Those customers will take a recommendation, and what tips them is the small stuff he mentions without thinking about it: this edible is vegan, that one has no artificial dyes. That's the customer the app most needs to get right, and the details are tiny.

Knowledge goes in the engine, not on the screen

Peter knows a lot. Indoor versus outdoor flower. Whole flower versus trim in a pre-roll. Rosin versus distillate in edibles and vapes, where his view is that rosin is the cleaner product. Disposable versus cartridge.

If I put all of that on screen, I'd break the first rule he gave me. So it doesn't go on screen. It goes into the logic, and into the questions the app asks. Depth is there when a customer asks for it and never shows up by default.

Deals and the dull chat

Peter wanted the deals at the top of the home screen and bigger than they are now, and he wanted the contextual ones to pop. I agree, to a point. There's a line where a deals banner turns a knowledgeable guide into a coupon app. I'd rather use a subtle pulse than countdown copy and "act now" language, which feels wrong for a nervous first-timer.

He also called the chat interface boring and dull next to the home screen. The home screen is blue and orange and high contrast, and the chat was navy and white. He was right. If you read the UI post, this is where that started.

What I took away

The overriding mission: K.I.S. Keep It Simple. (I left off the second S on purpose. No one is stupid.)

There's a lot of knowledge, nuance and guidance in a weed conversation, which underscores why a model is right for this project rather than just a POS menu on a tablet. Customers are curious. Even if they are savvy, there are so many new and innovative products being introduced that it necessitates a "cannabis sherpa" to offer assistance so they leave with their perfect product.