What Should I Use AI For First in My Business?
Start with the small thing that annoys you
On armchair ordering, twenty three recipes, and why my son now knows what he's cooking on Wednesday.
For years I ordered our groceries from the couch.
Mark called it armchair ordering.
I'd sit down with my phone, open the Giant app, and go from memory. I know what we like and I've made these dinners a hundred times. So I'd tap through, fill the cart, hit submit, and feel pretty good about myself.
Then Wednesday would roll around and I'd be standing in the kitchen with everything for the recipe except the one thing that made it that recipe.
I hadn't forgotten the whole shop, just the one thing. Then I'd throw together whatever I could out of what we had.
The other thing that kept happening was the complaining. We'd land on a handful of dinners that worked and eat them into the ground, and by week three somebody at the table would say something about it. Fair enough. They weren't wrong.
The routine took me about half an hour. Pull up recipes, work out what we needed, walk the fridge, walk the pantry, place the order.
Half an hour, and I'd still get it wrong on Wednesday.
This week I planned six dinners and placed the order in ten minutes. If even that.
What I actually built
It's a playbook, and AI runs it.
I wrote down the job and how I decide, and then I handed that over. Now I don't plan anything. It asks me a few questions about the week, gives me the dinners, and builds the order list. I say yes or I change something.
Here's what I put in it.
1. Every recipe we actually eat. About two dozen. Some had a card or a link. Others only lived in my head and I'd never written them down for anybody, including me.
Writing those out went fast. What it meant is that they finally exist somewhere besides my head, so now the system can work with me on them.
2. The make or break ingredient for each one. This was the whole reason I started. The item that, if it's missing, you are not making that dinner.
I didn't do this part. I loaded the recipes in and AI built a spreadsheet out of them, one row per recipe, and worked out which ingredient was the key one. It also sorted what could be swapped and what would be nice but nobody's driving to the store for. All I did was read it over and fix the few it got wrong.
3. Our rules. More fish. Red meat once a week, no more than that. Cauliflower crust pizza is Thursday and it doesn't move unless I say so.
4. Noah's restrictions, written out properly. He has Celiac disease and a peanut and tree nut allergy. That's medical, not preference, and it used to live entirely in my head while I shopped.
Now every recipe carries what to check on a label and whether it works on a night he's home. The playbook narrows the field and tells me where to look. I still make the call on what's safe, because that call is mine and it always will be.
That one turned out to matter more than the groceries did. I'd been doing that part from memory too.
5. Favorites, weighted. I went down the spreadsheet and marked the ones we're happy to eat often, right there in the cells. Those come up more. The rest are a once in a while thing, and now the playbook knows the difference.
6. What we ate last week. It keeps track, so nobody has to sit through the same six dinners on a loop. That's the complaint I was getting and now I don't get it. If I want a repeat, I ask for one.
7. What happens when people come over. Headcount, which recipes stretch, which ones don't.
That took me one afternoon to build, and most of it was typing out what I already knew.
What changed
Thirty minutes down to ten. That's the part people notice first, and it's the least interesting one.
Our grocery bill went down, because I'm not ordering to order anymore. No more grabbing things that seem like a good idea while I scroll.
I have everything I need for the meals I meant to make, on the day I meant to make them.
Nobody's complaining about eating the same dinner for the third week running, because we aren't.
We eat more fish, and red meat shows up once a week instead of whenever it felt like it. I didn't have to be disciplined about that. The rule does it for me.
Mark's cholesterol was down at his last checkup. I'm not going to hand a grocery playbook all the credit for that, but it sure didn't hurt.
And my favorite part. Noah is cooking. He knows which night is his before the week starts, so he's not getting asked at six o'clock. He's learning recipes he'd never made. That was never on my list when I sat down to fix a grocery order.
It's just like this in our business
We all do this with the parts of the business we know cold. You've run the thing enough times that you stopped writing anything down. It's in your head, it's fine, you'll remember.
And you mostly do remember. Then one day you forget, and of course it's the piece you needed.
I work with course creators, membership owners, and group coaches, so let me get specific about where I watch the armchair show up.
A new member joins and the one email that sets expectations doesn't go out. It isn't written down anywhere, it's just something you usually remember to do.
You get on a coaching call and can't recall what you promised her last month.
You've launched four times and rebuilt the whole sequence from scratch on all four.
Somebody asks a question in the community and you write a beautiful answer. Four months later somebody asks it again and you write it beautifully again, because the good answer is nowhere anybody can find it.
You are the file cabinet, and that's why it all routes through you.
Why I started at home
I did not build my first playbook for a client. I built it for dinner.
When people decide to get serious about AI, they reach for the biggest thing on their plate. Usually a launch, with money attached and a dozen ways to go wrong.
Then it doesn't work on the first pass. They can't tell whether the problem is the tool or the way they explained it, so they quit and decide AI is oversold.
Groceries were perfect for learning, because I could tell instantly whether it worked. Did I have the salsa on Wednesday or not? Nobody has to interpret that answer.
Pick something where you can look at the result and know in five seconds if it's right. That's what makes a good first one, and low stakes is the point, not a compromise.
How to tell if something should be a playbook
Rachel Woods teaches four questions for this, and they're the ones I use now on my own work and on client work. Take whatever job you're thinking about and ask:
How often do I do this? More often is better. A weekly job beats a twice-a-year job every time.
How much does it annoy me right now? Slow counts. So does doing it a different way each time and getting a different result.
Could I teach a new assistant to do it in five to seven steps? If you can teach a person, you can teach AI. If it takes thirty steps, the job is too big and needs cutting down first.
Do I know a good result when I see it? If you can picture what right looks like, you'll be able to tell whether the playbook is working. If you can't, you'll never know.
Now run dinner through all four. I did it every week and it bugged me every time. I could walk somebody through it in about six steps. And I know a good week of dinners when I see one.
Four yeses, so that was the one to start with. I didn't have to be clever to work that out.
Your goal here is picking the job that changes how your week actually goes, which is usually not the most impressive job on your list.
The part that took the work
Writing the steps was easy. An hour, maybe.
Writing down how I decide took the rest of the afternoon. Which recipes are heavy enough that we shouldn't have two back to back, and what "we've had this recently" actually means. Some ingredients can be swapped and some can't. I also had to say when to come ask me and when to just go ahead.
None of that was written anywhere anybody could use. Twenty years of making dinner, and it was all in my head or scribbled on the back of something, a scrap of paper, a page in a notebook I'd have to go hunting for.
I didn't know I knew it until I sat down to explain it to something that didn't.
That's the fear I hear from people all the time, by the way. My work is too specific for AI. It would get it wrong.
Nine times out of ten the tool is fine. What it's missing is everything you know that nobody ever wrote down properly. A brand new assistant would get it wrong for the same reason, and you'd never blame her for it.
So the job in front of you is getting your thinking out of your head and onto a page. That written thinking is what Rachel calls a playbook, and it's what I mean when I talk about getting your business to run without you.
Here's how I know that part is the asset.
I gave the playbook to my aunt. She cooks for one, so we changed things. Portions, obviously. Different rules, because her week doesn't look like mine and her kitchen doesn't either.
That took us almost no time at all, because the thinking was already down on paper. We weren't starting over, we were adjusting.
It's running now. She's eating at home more and eating better, and she never had to build a thing.
Once your thinking lives somewhere other than your head, somebody else can pick it up and make it theirs.
Go small on purpose
The trap is that once you start pulling on this, you see how many of these jobs you have. You don't have one playbook, you have twenty.
Then you think about every exception you've never written down, and it starts to feel impossible. So you either quit, or you go build one enormous thing that tries to cover everything and breaks on step two.
Mine doesn't cover everything. It does dinner. No lunches, no holidays, and nothing at all about what we eat when we're traveling.
It only has to be better than doing it by hand.
Ship it before it's finished. Use it for a week and let it show you what's broken, because that teaches you more than another week of planning. Fix that one thing. Then start the next job.
Five small wins over a month and a half will take you further than one perfect build you're still fiddling with at Thanksgiving.
And after a few of them the bigger projects stop looking so big, because you've learned how to cut a thing into pieces. That's the same momentum I mean when I say the work is feeling heavier than it should.
Done is better than perfect.
What the ten minutes is actually for
Hours saved is a small number. It's the one everybody reaches for and it's never the point.
Sunday used to have a chore parked in the middle of it that I'd put off and then do badly. Now it takes ten minutes and I'm out. Noah cooks Wednesday and learns something while he does it. We eat better without anybody white knuckling it.
That's what a playbook buys you. Your attention back, to spend on whatever you'd rather be doing. For me that's a walk in the woods with Mark, or curling up with a good book, or daydreaming about where we're going on the next trip.
It's the same question I ask clients before we build anything: what are you trying to be more present for?
Where I see this with the people I work with
Almost everybody I work with is good at what they do, and that's usually the problem. Being good at something means people bring you more of it, and pretty soon your week is full of work you can do and nobody stopped to ask whether you should.
So I come in and watch how the work actually moves. I talk to people, I look at the numbers, and I find the jobs eating the owner's attention.
Then we pick one and build it so it comes off her plate for good. I call it an AI Quick Start.
Dinner was mine. Yours is sitting on your calendar right now.
Here's what I'd do this week
Think of the small job you do from memory every week. The one that's fine until it isn't.
Run it past the four questions. How often, how much it annoys you, could you teach it in about six steps, and would you know a good result when you saw one. If you get four yeses, that's your first one.
Then write down three things about it:
The steps, in order, the way you'd tell a new assistant on her first day.
The make or break piece. The thing that, if it's missing, the whole job falls over.
How you decide. Exceptions, swaps, and the times you'd want somebody to come ask you instead of guessing.
That third one is the job. Give it the afternoon. Everything else is typing.
I put the four questions on one page for you, with room to write your answers. Grab it here.
So let me ask you. What's the thing in your business you've been armchair ordering, running off memory because you've done it a hundred times? Tell me in the comments. I'd like to know if yours is a Wednesday problem too.
Shell, Yes! Here's to your success,
Michelle
FAQ
-
A playbook is the written version of a job you do over and over. It holds the steps, the rules you follow, and how you decide along the way.
Most people write the steps and stop there. The deciding part is what makes it useful, because that's the piece that only lives in your head. Once it's written down, AI has something solid to work from instead of guessing, and so does any person you hand the job to.
-
Pick a job you already repeat, then run it past four questions. How often do I do this? How much does it annoy me right now? Could I teach a new assistant to do it in five to seven steps? Do I know a good result when I see it? Four yeses means start there. The right first job is usually small and boring, not the impressive one, because you need to be able to tell at a glance whether it worked.
-
Personal is a fine place to start, and often better. A home job has low stakes and a clear answer, so you find out fast whether the thing works and why. My first one planned our dinners and built the grocery order. What I learned building it is exactly what I use on client work, and I learned it without a launch on the line.
-
Usually because it doesn't have your context. You know the exceptions and what good looks like, and you've never had to say any of it out loud, because the job has never left your hands. Without that, you get the generic answer, which is the thing everybody complains about. Write down how you decide before you hand anything over.
-
Mine took an afternoon, and I was working from a job I'd done for years. The steps go fast. Writing down how you decide is where the time goes, and it's the part worth paying for. If it's taking you more than a day, the job you picked is probably too big. Cut it down to the smallest version that would still help you.
-
Find the work routing through you because it's in your head and nowhere else. That's different from work that needs your judgment. Write one of those jobs down, get it running without you, and repeat. I wrote about what that looked like the first time I stepped away from my own business for eleven days. Nobody's asking you to disappear. The aim is a week spent on the few things that actually need you.