ChatGPT for HR: Drafting JDs, Policies and Employee Letters Without the Usual Mistakes
Short answer: ChatGPT is useful for HR drafting: job descriptions, interview question banks, first drafts of policies, offer and warning letters, and internal announcements. It saves the blank-page time. It does not know your state's labour rules or your company's history, so every policy and letter needs a human check, and anything legal needs an employment lawyer's review.
Most HR teams in India already use AI quietly, usually for a JD or a festival announcement. The trouble starts when a draft goes out unchecked: a job post with an age limit, a leave policy copied from a US template, or a warning letter that reads like a threat.
This guide covers the four HR jobs where ChatGPT (or Claude or Gemini, which behave similarly for this work) genuinely helps, with prompts you can copy, sample outputs and what we would change in each before it goes anywhere near an employee.
What can HR professionals actually use ChatGPT for?
Use it for first drafts and structure, never for final decisions. In HR, that splits cleanly into work AI does well and work it should not touch.
| Good fit for AI drafting | Needs a human or lawyer, not AI |
|---|---|
| Job descriptions and job post variants for LinkedIn, Naukri and WhatsApp | Deciding who to shortlist, hire or reject |
| Interview question banks by role and level | Termination decisions and full-and-final calculations |
| First draft of a leave, WFH or travel policy | Checking a policy against state law and the labour codes |
| Offer, confirmation and appreciation letters | Responding to a legal notice or a labour inspector |
| Town-hall notes, festival and holiday announcements | POSH complaints and inquiry findings |
Screening resumes with AI is a separate question with its own bias and consent issues; our guide to AI resume screening covers it. For where HR fits in a wider office rollout, see how to use AI at work.
How do you write a job description with AI that avoids discriminatory wording?
Give the AI the real role, the reporting line and the must-have skills, and tell it explicitly what not to include. Without that instruction, models often copy the habits of old Indian job posts.
Say you are hiring a front-desk executive for a dental clinic in Janakpuri. A lazy prompt ('write a JD for a receptionist') can return lines like these, all of which we would strike:
- 'Female candidates preferred' or 'smart young girl': gender and age filters with no link to the job.
- 'Age 20–28 years': an age band shuts out capable people for no job-related reason, reads as discriminatory and shrinks your pool.
- 'Good-looking, presentable personality': appearance language is vague and easy to read as bias.
- 'Native Hindi speaker': ask for the actual need, such as 'able to speak with patients in Hindi and English'.
- 'Unmarried' or 'no family commitments': marital status has nothing to do with the work.
A better prompt: 'Write a job description for a front-desk executive at a 3-chair dental clinic in West Delhi. Duties: appointment calls, billing on our clinic software, patient reminders on WhatsApp. Must-haves: 1 year of front-desk experience, basic computer skills, Hindi and English. Shift 10 AM–7 PM, six days. Salary range ₹18,000–24,000 a month. Do not mention age, gender, marital status, appearance, religion or caste. Keep it under 250 words, with a short About the clinic line.'
The salary figure here is only part of the scenario; use your own band. Stating a range in the JD tends to cut irrelevant applications.
Can ChatGPT draft an HR policy for an Indian company?
It can draft the structure and the plain-language wording, but it cannot tell you whether the numbers are legal in your state. Leave, working hours, holidays and overtime in shops and offices are governed partly by each state's Shops and Establishments Act, plus central laws and the labour codes, and the details differ from Delhi to Maharashtra to Karnataka.
Here is how we run a leave-policy draft without letting the AI invent law:
- Write down your current practice in bullets: earned leave, casual leave, sick leave, carry-forward, encashment, notice for leave, who approves.
- Prompt: 'Turn these bullets into a leave policy for a 40-person private company in Delhi. Use simple English, numbered clauses, and a section for definitions. Do not add any legal entitlements I have not listed. Where state or central law may set a minimum, add the note VERIFY WITH COUNSEL.'
- Read the output and remove anything you did not supply. Models like to add entitlements you never offered, such as 'bereavement leave of 3 days', and US concepts such as 'at-will employment', which has no place in an Indian policy.
- Open your state's labour department page and the text of your state's Shops and Establishments Act (or the rules under the labour codes, as notified for your state) and compare each number. Minimum leave and working-hour rules are where drafts usually go wrong.
- Send the clean draft to your employment lawyer or labour-law consultant for review before it goes into the handbook.
- Only then ask the AI for a one-page employee summary of the final, approved policy.
The VERIFY WITH COUNSEL note is deliberate: it makes the gaps visible so nobody assumes the AI checked them. The Maternity Benefit Act, POSH obligations and gratuity rules are other areas where we have seen AI drafts state the law confidently and incompletely, so treat any legal statement in an AI output as unverified.
A sample HR letter, critiqued
Letters are where AI saves the most time and causes the most avoidable friction. Here is a typical exchange for a late-coming warning letter.
- Prompt: 'Write a warning letter to an employee who has been late 9 times in October.'
- AI output (opening line): 'This letter serves as a formal and final warning regarding your persistent and unacceptable tardiness, which will result in immediate termination if repeated.'
- Problem 1: 'final warning' and 'immediate termination' may not match your standing orders or your disciplinary process, and can create trouble if challenged later.
- Problem 2: no facts. A good letter lists dates and times, the policy clause, and what happens next.
- Problem 3: tone. It reads as an accusation, not a record of a conversation.
Better prompt: 'Draft a first written warning about late attendance. Facts: reported after 10:15 AM on these 9 dates in October (the dates follow); policy clause 4.2 says reporting time is 9:30 AM with a 15-minute grace. We discussed this verbally on 3 October. Tone: firm, respectful, factual. State that further delays may lead to action as per company policy. Ask the employee to reply in writing within 3 working days. No legal threats.' Then list the actual dates. The second draft is usually usable after light edits.
Replace the employee's name with 'the employee' while drafting, then add the real name in your own document. That habit keeps personal data out of chat history.
ChatGPT prompts for HR you can reuse
Save these as templates. Each one tells the model the context, the format and the limit, which is what separates a usable draft from a generic one.
- Interview questions: 'Give 10 interview questions for a mid-level accounts executive (Tally, GST returns, bank reconciliation). For each, add what a strong answer includes and one red flag. No trick questions.'
- Onboarding plan: 'Create a first-week onboarding checklist for a new sales executive at a 25-person distribution company. Day-by-day, with who owns each task.'
- Announcement: 'Write a 120-word WhatsApp announcement that the office will close early at 4 PM on the day before Diwali. Warm, simple English, no emojis.'
- Policy summary: 'Summarise this approved WFH policy into 6 bullets an employee can read in a minute. Do not add anything that is not in the text.'
- Exit interview: 'Draft 8 neutral exit-interview questions that encourage honest answers about manager, workload and pay.'
Performance reviews need their own approach because the risk is vague, inflated feedback; see our guide to writing performance reviews with AI.
What should HR never paste into ChatGPT?
Do not paste anything you would not put in an email to a stranger: Aadhaar or PAN numbers, salary slips with names, medical records, POSH complaint details, or disciplinary files. Even on business plans that say they do not train on your data, the safer habit is to redact first.
Agree the rules in writing before the team starts. A short company AI usage policy that names allowed tools, banned data types and who reviews outputs prevents most of the problems we see. To check a tool's data handling, read its current privacy and data controls page; settings change, so recheck them every few months.
Where to build the skill
Most HR teams get comfortable in two or three weeks of daily use on their own drafts. If you want structured practice across ChatGPT, Claude, Gemini and NotebookLM (for long policy documents), our Applied AI Professional course covers prompting for office work with no coding.
Frequently asked questions
Is it legal to use ChatGPT for HR work in India?
Using AI to draft documents is not illegal in itself. The legal risk comes from what you paste in (personal data of employees) and what you send out unchecked. Redact personal data and have policies and legal letters reviewed by a qualified person.
Can ChatGPT write an HR policy that complies with Indian labour law?
Not reliably. It does not know your state's current rules or recent amendments, and it sometimes states law that does not exist. Use it for structure and wording, then verify every entitlement with the official source and an employment lawyer.
What are the best ChatGPT prompts for HR?
The best prompts give the role or situation, the exact facts, the tone, the format and a limit on what to add. Prompts that say 'do not add anything I have not listed' prevent most invented clauses.
Will AI replace HR professionals?
AI is taking over drafting and first-level answering of routine policy questions. Judgement calls, employee relations, investigations and legal compliance still need people, and those parts of the job are becoming more important, not less.
Which AI tool is best for HR: ChatGPT, Claude or Gemini?
For drafting, all three are capable. Choose based on what your company already uses (Gemini sits inside Google Workspace, Copilot inside Microsoft 365) and on the data controls available on your plan, as of November 2026.
Keep reading in AI at work
- How to Use AI at Work: Practical Examples for Every Role in an Indian Office Start here
- ChatGPT for Performance Reviews: From Your Notes to a Fair, Specific Appraisal
- Corporate AI Training in Delhi NCR: What to Check Before You Book
- ChatGPT for Excel Formulas: Worked Examples on GST, Salary and Sales Sheets
- AI Training for School Teachers: How to Plan Sessions That Change Classroom Practice
Courses: Applied AI Professional · Corporate AI training · All guides in AI at work