Skip to content
Sam Jennings
← ALL FIELD NOTES
009 · 2026-11-13 · OPERATORS & AI · 4 MIN

You don’t need an AI strategy. You need a list.

In June I told operators to skip the strategy and write down every task they repeat. “Make a list” is easy to agree with and hard to start, so here’s a worked one, from five years running a taproom.

In June I wrote that the first step with AI isn’t a strategy. It’s a list. For one week, write down every task that’s repetitive, annoying, or typed more than once.

People agreed with that, and a couple saved it. I’d bet most of them never made the list. Fair enough. It’s easy advice to nod along to and hard to sit down and do. So here’s a worked one.

I ran the taproom at Camp Beer Co. for about five years. This is the week I’d have written down, near enough:

THE WEEKHOW OFTEN
rebuild the menu PDF when a keg blowsevery keg change
reply to private booking enquiriesmost days
answer the same questions from new staffevery new hire
write next week's scheduleweekly
reply to reviewsa few a week
turn the sales report into a summaryevery Monday
chase a supplier about an orderweekly
post the new release on Instagramevery release

Nothing on it is interesting. That’s the point. Interesting work doesn’t repeat.

Now the step most people skip. Not everything on the list is an AI job, and sorting it before you touch a tool is where the real value is. Every line goes in one of three piles.

Some of it is a software problem. The menu was the biggest item on my list, about three and a half hours a week, and AI was never the fix. The menu was hard to change because it lived in three places at once. The fix was giving it one source of truth, which is a different kind of project entirely. I wrote that one up in an earlier note.

Some of it is a process problem. The new-staff questions didn’t need AI to answer them. They needed someone to write the answers down once. AI can help you draft that document in an afternoon, but the job was always the document.

And some of it is genuinely an AI job. The output is words, the inputs already exist, and a person can check the result in a minute. On my list, that’s three.

Booking enquiries. The facts are the same every time: capacity, minimum spend, which dates are open, what’s included. Only the wording changes. Write the facts down once, and AI drafts every reply for a manager to check and send.

Review replies. A draft that thanks the person for the specific thing they mentioned, in the taproom’s voice, for someone to approve. Nobody should let these post themselves.

The Monday summary. Export the sales report, hand it over, and ask what changed from last week and what’s worth telling the owners. Spreadsheet squinting becomes a read-through.

Not everything annoying is an AI job. Sorting the list tells you which half is.

Look at what those three have in common. Defined inputs. A defined output. A person checking before anything leaves the building. That’s what I meant in June by a defined process, and notice that I didn’t have to invent one. Sorting the list defined them for me.

The rest of the list isn’t wasted either. The software pile and the process pile are your next two conversations about what to fix, and they’re usually worth more than the AI pile.

So, the same homework as June with one extra step. Write the list for a week. Sort every line into software, process, or AI. Hand over the three most annoying AI jobs. Then look hard at the other two piles, because that’s where the hours are hiding.

Working on something like this? hey@samjennings.dev