Ask it to list its assumptions first.
Silent guesses feel like answers. Make the model show them before it commits.
Braxton Ellsworth
AI Systems Architect
If you ask AI to list its assumptions before it answers, you catch the silent guesses while you can still fix them. A finished reply can look clean and still be built on a wrong default. By then you're editing the polish instead of the foundation.
Prompt Fluency is communicating so the model doesn't have to guess. Here's a move that flips the guessing: ask it to show the blanks it was about to fill, then fill them yourself.
Picture this. You run a print shop. You type:
"Write a confirmation email for a rush business card order."
You get a friendly message. It assumes 500 cards, next-day pickup, a card stock you don't carry, and a tone that sounds like a big corporate print chain. None of that is crazy. None of it is your shop. You spend ten minutes rewriting a message that never should have invented those details.
Now try it again with one line up front:
"Before you write the email, list every assumption you're about to make. Number them. Wait for my answers. Don't draft yet."
It comes back with:
- Quantity
- Pickup or shipping
- Paper stock and finish
- Due time
- How formal the customer expects
You reply in thirty seconds. Then it drafts. The email matches the order because the guesses never got a chance to dress up as facts.
Same thing shows up everywhere.
A pet boarding desk asks for a weekend check-in checklist, and the model assumes dogs only, then skips the cat meds and the late pickup fee.
A high school band booster asks for a fundraiser flyer, and the model assumes a bake sale when the real event is a car wash with a rain date.
An apartment office asks for a pool hours notice, and the model invents adult-only evenings that aren't in the lease.
Each time, the finished draft looked helpful. The wrong assumptions were already baked in.
So ask for the list first. Plain English.
"Before you write anything, list the assumptions you're making about quantity, timing, tone, and what 'rush' means here. Wait for me."
"List what you're assuming about the animals, the med schedule, and pickup rules. Don't draft the checklist until I confirm."
"Before the flyer, list assumptions about the event type, date, price, and who it's for. Number them."
"Before the pool notice, list assumptions about hours, age rules, and guest policy. I will correct them."
Look at what changed. You stopped treating the first draft as the starting point. You made the model show the invisible fill-ins. Then you replaced them with the real ones.
Think of it like a contractor pricing a kitchen. If they start swinging a hammer before they ask about the floor, the cabinets, and whether the wall is load-bearing, the pretty work sits on a wrong plan. A short assumption list is the walk-through before the build.
No shame here. I skip this too when I'm in a hurry. I'll ask for the email, the checklist, the flyer, and then get mad at a polished answer that quietly chose for me. The model didn't sneak anything past me. I never asked it to show its work.
A quick check before you ask for a draft:
Ways to ask for assumptions first
- Say "list your assumptions" and "don't draft yet"
- Ask for a numbered list so you can answer line by line
- Name the areas that usually go wrong (timing, quantity, tone, audience, rules)
- Reply with corrections, then say "now draft"
- If the list is short and vague, ask "what else are you filling in that I didn't say?"
When a finished answer feels almost right but wrong in three places, ask yourself: did I make it show the guesses first? Odds are you only asked for the deliverable.
You're not slowing the chat down. You're removing the need to guess. That's Applied Intelligence Mastery in practice: surface the assumptions, lock the real ones, then let the model write from a plan you both can see.
Want to think in systems, not prompts?
Take the free AIIQ test to measure your AI fluency, or enroll in the full Applied Intelligence Mastery program.