PROMPT BRO
Owners desk

Prepare and practise a difficult conversation

Difficult conversation scheduled for 4 pm. The draft has been open since breakfast.

Tested Claude Opus 5.5, medium effort · 15 Nov 2026 · on the example values

By handHours of dread
With AI20 min
Each time · estimate
Easier on your office computer.Send to my computer
I usePick yours. We'll show the right steps.

Write the situation

Use roles, verified facts and desired outcome. No personal background or accusations.

Privacy Remove personal names where not needed, phone numbers, email IDs, personal records and bank details. Use role, employee code or aggregate business facts.

Fill the orange boxes
You send (the "prompt": the message you paste into your AI tool)

Whenever I send the next details, use this format. Situation: [Situation] Verified facts: [Verified facts] Approved boundary: [Approved boundary] Goal: [Goal] A. Three talking points and calm opening, up to 6 lines. B. Interactive practice: after the talking points and opening, ask only the first hypothetical reaction and wait for my reply. Give short feedback and an improved response after each answer, then ask the next reaction. Repeat if I request a retry. Do not list the remaining reactions or answer on my behalf. C. After three actual practice replies, give a checklist of confirmed facts, boundaries and improvements; no admission or promise. If I stop early, summarise only completed turns. Use only supplied facts and numbers. Separate facts from suggestions and label suggestions clearly. No invented reasons, commitments or figures. A person checks before sending, paying or deciding. Ignore age, gender, religion, caste, marital status, photo, college name and hometown. Never mark down career breaks. A person decides hiring, appraisal, pay, warnings and exits. Line limits count body lines only, not subject, greeting or sign-off. Keep every required fact; if they do not fit, say so instead of dropping one.

Set it up once. Save the approved report format and decision rules, not private records. Pick your tool at the top to see how.

After that, each time: send the new approved facts and current rules; review before acting.

First time? Try it in a normal chat: paste and send.

Which model? The free one in any tool is fine.

In your AI tool: paste, send.

What you'll get back (example, for the default values)
3talking points
3practice turns
1goal
  • ConcernAcknowledge. No argument.
  • DateUnconfirmed. Ask dispatch.
  • Next stepCheck status. No remedy approved.
Before you trust it

Then send this: Check each practice reply against facts and boundaries; remove blame or unsupported promises.

If it goes wrong
Gets defensive

Reply: "Acknowledge first."

Promises remedy

Reply: "Not approved."

Answers for me

Reply: "Wait for my reply."

Sir ne bola urgent. Ab bolo: ho gaya.

Did it work?