Say which tools you actually have.

If you say which tools you actually have before you ask AI for a process, checklist, or workflow, it stops inventing steps that need software you don't own. A clean SOP can look sharp and still be useless if every other line assumes an app, export, or gadget that isn't in your shop.

Prompt Fluency is communicating so the model doesn't have to guess. Here's a move that cuts the fantasy: put the real toolkit in the prompt so it builds for your bench, not a demo kitchen.

Picture this. You run a neighborhood coffee shop. Closing keeps slipping, so you type:

"Write a closing checklist for our coffee shop."

You get a polished list. It tells you to export the day's sales from Square, reconcile Toast tips, clear the tablet queue queue, and push a Slack handoff to the opener. Your shop has a paper clipboard by the espresso machine and Google Sheets on one shared phone. None of those fancy steps are crazy. None of them are yours. You just spent ten minutes reading a fantasy SOP.

Now try it again with the toolkit named up front:

"Write a closing checklist for our coffee shop. Tools we actually have: a paper clipboard, a shared Google Sheet on one phone, cash drawer, and a basic POS that only prints a Z report. No Square export, no Toast, no tablet app, no Slack. Every step has to work with what we have."

It comes back with clipboard lines, a phone photo of the Z report, a Sheet row for cash vs. report, and a sticky note for the opener. Same job. Real tools. Something a closer can finish at 8pm without hunting for software you never bought.

Same thing shows up everywhere.

A bike shop asks for a repair-status text to customers, and the model invents Shopify notifications and a CRM drip. They text from one shared Android.

A food truck asks for a weekly prep list, and the model invents a restaurant POS inventory module. They use a cooler whiteboard and a Notes app.

A PTA lead asks for a spirit-wear order form email, and the model invents a Canva Brand Kit plus Mailchimp automation. They have Gmail and a Google Form.

Each time, the finished draft looked helpful. The steps needed tools nobody on the team can open.

So name the toolkit. Plain English.

"Tools we have: paper clipboard, Google Sheets on a phone, cash drawer, basic POS Z report. No Square, no Toast, no Slack. Write the closing checklist for those only."

"We text customers from one shared Android. No Shopify alerts, no CRM. Write a short repair-status text we can copy and send by hand."

"Prep tools: cooler whiteboard and iPhone Notes. No POS inventory module. Write a weekly prep list we can rewrite on the board."

"We have Gmail and a Google Form. No Canva Brand Kit, no Mailchimp. Write the spirit-wear order email and point parents to the Form link."

Look at what changed. You stopped asking for "a process" in the abstract. You told the model which bench it's standing at. Then the steps matched the gear.

Think of it like asking a friend for a dinner plan. If you don't say your kitchen has a skillet, a microwave, and no oven, they'll send you a roast. The plan wasn't wrong for a fancy kitchen. It was wrong for yours.

No shame here. I did this too. I'll ask for the checklist, the customer text, the prep list, then get mad at a polished SOP that quietly rented software I don't have. The model didn't sneak past me. I never said what was on the shelf.

A quick check before you ask for a process:

Ways to name the tools you actually have

• List the apps, paper, phones, and gear you really use
• Say what you do not have (exports, CRMs, tablet apps, automations)
• Force every step to run on that short list
• If a step needs a login you can't open, cut it or replace it
• When the draft feels "enterprise," ask "rewrite using only the tools I named"

When a workflow looks almost right but needs three apps you don't run, ask yourself: did I say which tools I actually have? Odds are you only asked for the checklist.

You're not limiting the model. You're removing the need to guess. That's Applied Intelligence Mastery in practice: name the real toolkit, kill the fantasy steps, then let the model write a process you can run tonight.