I've had 200 conversations about AI with food and drink founders. Almost everyone is stuck in the same place.

Over the past year I’ve had conversations with more than 200 food and drink founders and operators about AI.

Almost all of them said a version of the same three things.

One founder described AI to me as “an ambitious but naive intern” - which is a pretty good description if you ask me.

The content out there doesn’t help much either. Most AI content for food and drink is either an enterprise case study about productivity gains on PepsiCo’s factory floor, or an AI-fluencer selling a course on the one prompt that supposedly made them £100M. Neither is any use if you’re a small team trying to work out whether AI can help you fill in a new line form faster.

So that’s what this is for. Every week I’ll take one thing I’ve learned from those conversations, or from the brands we work with at Pesto, and write it up properly. No tool round-ups. No predictions about 2030.

I co-founded an Indian ready meal brand called Tuk In. Alongside my co-founder Tom we brought the curry-in-a-naan from kitchen table to retail and foodservice.

I’ve got the battle scars - going door to door to convenience shops trying to convince them to stock our product, then hand stocking the shelves when I got one to agree. Signing up to Lunch! trade show 2 weeks before with no marketing stand and no product and scrambling to get it done in time (that’s the photo above, me looking much more youthful and with our prototype branding). Delivering 14,000 naans to 12 airports around the country with no batch codes on and then driving round them all and sitting in industrial freezers hand applying labels to every single one.

After Tuk In I spent four years at Bain, advising businesses on technology strategy and how to actually improve their operations and margin. It’s where I learned that most transformation programmes fail for deeply unglamorous reasons, usually to do with data nobody wanted to tidy up. The stuff that can be done with AI now was literally a junior consultant’s job for an entire 12-week project.

Now I’m COO and co-founder at Pesto, where we’re building AI-native operations software for food and drink. We work with brands like Punchy, Bold Bean Co, The Curators and Well&Truly to help them transform the way they run their businesses using technology.

AI does six different types of work. Most brands only use one of them.

Have a think about what your business actually uses AI for right now. If you’re like nearly every brand I speak to, the answer is writing things. Social captions, product descriptions, first drafts of buyer emails.

That’s one type of work. I call it Create.

The other five are Research, Evaluate, Interpret, Plan and Execute. Put them together and they spell RECIPE - sorry, not sorry.

Create is the one everyone finds first, because it’s the most visible and the least frightening. It’s also, for most food and drink businesses, pretty basic. The biggest opportunity sits in Interpret and Execute. Reading your own sell-through data properly before a range review. Getting a PO checked against your agreed pricing and generating the invoice correctly. One brand we work with lost £4k because they took the retailer’s word for it.

You don’t need to do anything with this yet. Just notice which type of work you’re using AI for. Over the next few weeks I’ll go through where the other five actually apply in a food and drink business, how to spot the opportunities you’re walking past, and how to work out which one is worth doing first.

One piece a week grounded in something real, a brand I’ve worked with, an insight I’ve seen directly or a conversation I’ve had. I’m speaking to brands every week and helping our customers solve their biggest challenges with AI.

Some weeks it’ll be a framework. Some weeks it’ll be me telling you about something that didn’t work.

What’s the one task in your business you’d hand over tomorrow, if you trusted something else to do it properly?

Hit reply and let me know - it’ll shape what I write about.