Steal My Prompt Vol. 41: The Objectives and Key Results Pressure-Tester

Every quarter you write Objectives and Key Results (OKRs) that look sharp in the doc. Clean objectives, measurable key results, the right amount of ambition. Then you present them, one leader asks one question, and you realize the goals were built to be approved, not to survive the quarter.

OKRs fail in three predictable ways. The objective sounds aspirational but is really just activity in a nicer outfit. The key results look measurable but carry no baseline, so nobody can tell if you actually moved anything. And the whole set quietly assumes you can fund everything at once, which you cannot. Most drafts never get tested against any of these before they lock for the quarter.

So I built a prompt that plays the one person you are actually writing for: the skeptical leader in the approval room. It reads your draft OKRs, asks a couple of sharp questions about your capacity and your audience, then attacks the goals the way that leader will. Relevance, measurability, ambition, and the tradeoff nobody wants to name.

The goal is defensible OKRs, ready before the meeting instead of during it. You walk in already knowing where the soft spots are, because you already lost that argument in private.

What You Can Use This For

• Pressure-testing your quarterly OKRs before you submit them for review
• Stress-testing a single objective you already suspect is weak but cannot explain why
• Preparing for the hard questions leadership will ask in the OKR review meeting
• Turning a vague key result ("improve onboarding") into a real one with a baseline and a target
• Deciding which key result to cut when your team's capacity does not match your ambition
• Reviewing a direct report's or a clinical operations team's OKRs before goals lock for the quarter

How to Use It

1. Open Claude, ChatGPT, Microsoft Copilot, or Gemini. This works on the free tier of all four.
2. For a real submission, turn on the reasoning model: Extended thinking in Claude, the reasoning model in ChatGPT, Think Deeper in Copilot, or Deep Think in Gemini. The pressure-test catches more on the first pass when the model reasons before it answers.
3. Copy the prompt below, paste your draft OKRs into the first field, and add your capacity and your reviewer if you have them handy. If you leave those blank, the prompt will ask.
4. Answer its questions honestly. The tradeoff question is the one that stings and the one that matters. If you cannot name what you would cut, your OKRs are a wish list, not a plan.
5. Revise, then run it once more. A second pass on a tightened draft surfaces the deeper problems the first pass hid.

Pro tip: paste last quarter's OKRs and your actual results first, and ask it to grade how well your goals predicted reality. You will see your own pattern (usually sandbagged key results, or objectives that were really just projects) and you will write the next set with your eyes open.

The Prompt

You are a skeptical senior leader reviewing my draft OKRs in a pre-submission

review. Your job is not to encourage me. It is to find every place these goals

will fall apart under scrutiny before the leader who approves them does.

Here are my draft OKRs:

[PASTE YOUR DRAFT OBJECTIVES AND KEY RESULTS]

My team's capacity and constraints this quarter:

[HEADCOUNT, COMPETING COMMITMENTS, AND ANYTHING ALREADY PROMISED. LEAVE BLANK IF UNSURE.]

Who reviews or approves these, and what they care about most:

[NAME THE ROLE AND THEIR TOP PRIORITY, FOR EXAMPLE A CFO FOCUSED ON COST OR A

VP FOCUSED ON GROWTH. LEAVE BLANK IF UNSURE.]

First, read my OKRs. Then check whether I gave you my capacity and my reviewer's

priorities. Ask me only for what is genuinely missing, one question at a time, no

more than three questions total, and stop as soon as you have enough. If I already

provided everything, skip the questions and begin the review.

Then run this review:

1. OBJECTIVES. For each objective, decide whether it is worth a full quarter to

the reviewer who approves it, or whether it is activity dressed as ambition. Say

which, and why.

2. KEY RESULTS. Put every key result in a table with four columns: the key result,

whether it has a real baseline and target (yes or no), whether it is a genuine

stretch or a safe bet you will clear easily, and the one change that would make it

stronger.

3. THE TRADEOFF TEST. Based on my capacity, tell me which key results cannot all be

fully resourced at the same time. Then make me choose: name the single key result I

would cut first if the quarter goes sideways, and pressure-test whether my reasoning

holds.

4. THE HARDEST QUESTION. Given who approves these, write the single hardest question

that reviewer will ask, and the honest answer I do not have yet.

End with a one-line verdict for each objective: submit as written, revise first, or

rethink.

Do not soften your feedback to be encouraging. I would rather lose this argument

with you now than in the room.

Transparency and Notes

• Built and tested in Claude and ChatGPT on the free tier. Works in Gemini and Copilot too. No paid feature required, though the reasoning model sharpens the critique.
• Model-agnostic. The adaptive triage (it asks only for what you did not provide) works in any of the four tools.
• This is the adversarial reviewer pattern: you assign the AI the role of your toughest audience and let it attack your work before that audience does. It transfers straight to business cases, roadmap reviews, budget requests, and promotion write-ups. Swap the role and the criteria, keep the structure.
• Privacy note: OKRs often carry confidential targets and headcount. If yours include sensitive numbers, generalize them or use a personal account rather than a work tool bound by data policies you have not checked.