Answer staff questions from your policy
Answering 'how many leaves do I have' for the 900th time
Tested Claude Opus 5.5, medium effort · 15 Nov 2026 · on the practice files
Get approved policy extracts
Put the current rules in Excel with a rule ID, topic and approved text. Use only the current version.
Remove personal details
Type generic questions without names, balances, claims or case details.
Privacy Remove staff names, employee codes tied to cases, balances, contact details and claims. Use approved policy extracts only if your company allows it.
Whenever I send you approved policy extracts and staff questions, answer only from those extracts. Policy version: [Policy version] Questions: [Staff questions] A. One short reply per question, up to 6 lines, with the relevant rule ID and wording. If no rule covers it, say "Not in the supplied policy. Please check with HR." B. Table: question, rule ID, answered / needs HR. C. List missing or conflicting rules for HR to clarify. Do not invent policy, entitlements, legal rules, balances or approvals. An answer is guidance, not permission. If dates or individual cases are needed, send them to HR. 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 instructions and approved policy extracts in your company's approved tool. Replace old versions when policy changes. Pick your tool at the top to see how.
What is that?
A Project is a folder in ChatGPT that keeps your instructions and files for every chat inside it. It's on the free plan too. Checked 14 Nov 2026.
What is that?
A Project in Claude is a workspace with its own instructions and files that every chat inside it uses. Checked 14 Nov 2026.
What is that?
A Gem is a custom Gemini that loads your instructions before every chat. Free on every plan; you create it on a computer. Checked 14 Nov 2026.
What is that?
Signed in with a work account, Microsoft's Copilot Chat applies enterprise data protection: your prompts aren't used to train its models. Checked 14 Nov 2026.
After that, each time: send the new generic questions; verify the rule before replying.
First time? Try it in a normal chat: paste and send.
Which model? Use your tool's strongest model with thinking turned on.
In your AI tool: attach the files with the + button, paste, send.
- P3WFH. Manager approval is required.
- P2Expense bills. Submit within 10 days of expense.
- MissingCarry forward / travel advance. Neither rule is supplied; ask HR instead of guessing.
Then send this: Match every answer to an exact policy line. List unsupported statements, correct them, and confirm questions with no rule remain unanswered.
It answers from general knowledge
Reply: "Use only the attached policy; mark missing rules."
Two rules conflict
Reply: "Quote both and ask HR which is current."
It grants leave or WFH
Reply: "Explain the approval process only."
Sir ne bola urgent. Ab bolo: ho gaya.
Did it work?
If your whole team does this, we'll set it up for all of them in one session.