Say which tradeoff you already picked.

A clear ask still drifts when the model has to invent your priority. Say which tradeoff you already picked in the AI prompt so it stops writing a soft middle that serves neither side.

Prompt Fluency is communicating so the model doesn't have to guess. A tradeoff is a choice you already made between two good things you cannot fully have at once. Firm or friendly. Fast or careful. Keep the sale or keep the margin. You picked one. The model did not hear the pick. So it splits the difference and gives you something that sounds balanced and lands nowhere.

Try this ask:

"Write a reply about the late tool return."

You get a note that is polite, a little firm, a little forgiving, and somehow also vague about the fee. That is not your policy. That is the average of every customer reply the model has seen. You already decided something. You just never said which side won.

The model didn't ignore you. It wrote a reply. It invented a compromise you never asked for.

Same break shows up everywhere.

"Draft the note about costume pieces for the school play."

"Write the stall message for Saturday morning at the market."

"Make the late-fee text for the self-storage office."

Costume note. Market message. Late-fee text. Every one of those hangs on a tradeoff you already settled in your head: reuse what you have or buy new, sell out early or hold stock for regulars, enforce the fee or waive it once. Leave that pick out and the model will invent a soft middle that protects nobody.

Name the tradeoff. Plain English. Right in the prompt.

"Write a reply about the late tool return at the hardware rental desk. Tradeoff already picked: keep the customer and waive the late fee this one time, but state the fee clearly so they know what happens next time. Do not invent a partial charge. Do not sound like collections."

"Draft the note about costume pieces for the school play. Tradeoff already picked: reuse the closet first, buy only what is missing after the inventory check. You can invent a calm coordinator tone. Do not push a full Amazon haul."

"Write the stall message for Saturday morning at the farmers market. Tradeoff already picked: hold two boxes of berries for the regulars who texted, sell the rest as they come. You can invent a friendly line. Do not promise unlimited stock and do not hide that some is held."

"Make the late-fee text for the self-storage front office. Tradeoff already picked: enforce the $25 late fee, offer a payment plan of two weeks, no full waive. You can invent clear wording. Do not invent a free pass."

Look at what changed. Same job. Same need for decent wording. But now the model knows which side of the fork is closed. It can still write. It just can't reopen a decision you already made.

No shame here. I do this too. Once the tradeoff is settled in your gut, it stops feeling like something you need to type. It feels like background. The model has no background. It only has the words.

A quick check before you hit send:

Tradeoffs that need a line in the prompt

• Firm vs friendly (and which one wins this time)
• Speed vs care (ship today, or double-check first)
• Keep the relationship vs keep the rule (waive once, or enforce)
• Quantity vs quality (more options, or one strong answer)
• If you would argue with a coworker about which side to pick, name the winner for the model

When the answer comes back sounding fair and still useless, ask yourself: did I say which tradeoff I already picked? Odds are the model split a difference you already settled. Name the pick. Leave the wording free. Ask again.

You're not being rigid. You're removing the need to guess. That's Applied Intelligence Mastery in practice: say which tradeoff you already picked, so the model writes to your decision instead of inventing a middle path.