Should AI make shortlisting decisions for HR in India?

No. Use AI to score resumes against written, job-related criteria and to explain each score, but keep a person responsible for every shortlist and rejection. Test the tool with paired resumes for bias, never feed it attributes such as caste, religion or marital status, and tell candidates in your notice that AI assists screening.

Do the new Labour Codes change HR policies?

Yes. The four Labour Codes took effect on 21 November 2025 and consolidate 29 central labour laws. Existing rules continue while new rules are framed, so review policies such as appointment letters, fixed-term contracts, gratuity and night shifts for women with counsel. Do not rely on an AI draft that cites the old Acts.

Is it safe to paste employee data into ChatGPT or Gemini?

Not into a personal account. Employee and candidate data is personal data under the DPDP Act, 2023. Use only your company's approved AI tool signed in with a work account that has data protection commitments, remove names and contact details where the task does not need them, and follow your retention schedule.

Which HR tasks should we try with AI first?

Start with low-risk, high-volume drafting: job descriptions, structured interview questions with rubrics, plain-language rewrites of existing policies, onboarding checklists and training quizzes. These save time quickly and a human reviews every output. Leave screening, rejections and disciplinary matters until your rules, bias tests and review process are in place.

Can an AI agent answer employees' HR questions?

Yes, if it is grounded in your current policy documents, cites the source section and hands over to a person for anything sensitive or unclear. Microsoft offers HR-focused templates such as the Employee Self-Service agent, but says it is not a substitute for decisions by HR or other qualified staff.

AI for HR Teams in India: Hiring, Policies and Training

AI can save Indian HR teams real time on job descriptions, interview kits, first drafts of policies, onboarding plans and learning paths. It should not decide who gets hired, and every policy it drafts needs checking against the four Labour Codes, in force since 21 November 2025, and the Digital Personal Data Protection Act, 2023. Use it as a fast drafter with a named human owner.

At a glance

Use AI for: job descriptions, structured interview questions and rubrics, policy drafts and plain-language rewrites, onboarding plans, training calendars and quizzes
Keep humans on: shortlisting decisions, rejections, disciplinary matters and final policy wording
Watch for: bias against career gaps, college tier, hometown or gender, and old law in policy drafts
Data rule: no candidate or employee data in personal AI accounts
Legal context: the four Labour Codes took effect on 21 November 2025; most DPDP Rules, 2025 obligations apply from May 2027

This article is general information for HR teams, not legal advice. Have counsel review policy wording before you publish it.

Hiring: where AI helps and where it hurts

Job descriptions

Give the tool the real role: team, reporting line, the five outcomes that matter in the first year, must-have skills, location and work pattern. Ask it to separate essential from desirable requirements and to remove phrases that put good applicants off, such as age hints ("young and dynamic"), gendered wording, or "immediate joiners only" when you could actually wait a month. Then read the draft aloud. If a requirement would not change your hiring decision, delete it.

Screening with care

This is where AI can do the most damage. A tool that ranks resumes learns patterns, and patterns can carry bias: marking down a two-year career break, which is common among women returning to work, favouring a handful of colleges or metros, or reading surnames as signals. Attributes such as caste, religion, marital status and family plans must never be inputs. Laws such as the Rights of Persons with Disabilities Act, 2016 are a reminder to check that screening does not shut out candidates who need reasonable accommodation.

Practical safeguards that work:

Criteria first. Write scoring criteria from the job description, and make the AI score against them with a stated reason for each score.
Paired tests. Run two resumes with the same skills, one with a career gap or a smaller-town college. If the scores differ, fix the criteria before you use the tool on real applicants.
Human sign-off. A person reviews every rejection that AI influenced, and the recorded reason is job-related.
Tell candidates. If AI assists screening, say so in plain words in your recruitment notice.

Interview kits

AI is excellent at structured interviews. For each competency, ask for two behavioural questions and a rubric describing strong, adequate and weak answers. Structured interviews are fairer than free-form conversations and far easier to defend later. Have the hiring manager edit the kit, because only they know what "strong" looks like in their team.

Candidate and employee data under the DPDP Act

HR holds some of the most sensitive personal data in any company. The DPDP Act, 2023 allows some processing without consent for "legitimate uses". Section 7(i) covers purposes of employment and safeguarding the employer from loss or liability. Candidates are not yet employees, so do not assume that ground covers recruitment; take advice on the notice and consent you give applicants.

Section 8(7) requires a data fiduciary to erase personal data when consent is withdrawn or once it is reasonable to assume the purpose is no longer served, unless a law requires it to be kept. Rejected candidates' CVs sitting in old chat histories and shared spreadsheets are a common gap.

On timing, MeitY notified the DPDP Rules, 2025 in November 2025 with phased commencement. Most operational rules, including notice, security safeguards and breach intimation, come into force eighteen months after publication, which is May 2027. Use the runway. Our article on the DPDP Act and AI tools goes deeper.

Four simple rules for now:

No candidate or employee data in personal AI accounts.
Use the company's approved tool, signed in with a work account. Microsoft, for example, states that Copilot Chat prompts and responses under enterprise data protection are not used to train foundation models.
Remove names, phone numbers and addresses when the task does not need them.
Set retention periods for recruitment data and delete on schedule.

Policy drafting after the Labour Codes

On 21 November 2025 the Ministry of Labour and Employment made four Labour Codes effective: the Code on Wages, 2019, the Industrial Relations Code, 2020, the Code on Social Security, 2020 and the Occupational Safety, Health and Working Conditions Code, 2020. Together they consolidate 29 central labour laws. The ministry's release says the existing provisions of the old Acts and their rules continue while new rules are framed, so many policies are in transition.

That matters for AI. Many tools were trained mostly on material written before the Codes and will draft a policy citing the Payment of Gratuity Act, 1972 or the Factories Act, 1948 as if nothing had changed. The ministry's release highlights changes your policies should reflect, including:

Mandatory appointment letters for all workers
Fixed-term employees getting benefits comparable to permanent staff, with gratuity eligibility after one year of continuous service
Women permitted to work night shifts, subject to their consent and required safety measures
First statutory definitions of gig work, platform work and aggregators

Where AI fits in policy work:

Policy	AI can draft	A human must check
Leave and attendance	Structure, plain language, FAQs	Code rules and your state's shops and establishments law
Fixed-term contracts	An explainer for employees	Gratuity and benefit terms under the Codes
POSH	Awareness material, training quiz	Internal Committee and process under the 2013 Act
Use of AI at work	A first draft	Approved tools, data rules, confidentiality

AI is also useful for translating policies into Hindi, Tamil, Marathi or other languages for shop-floor and field teams. Always have a fluent colleague check the translation, because a wrong word in a leave or safety policy causes real confusion.

Onboarding and training plans

Onboarding

AI can turn your policy documents into a 30-60-90 day plan for each role, a first-week checklist and a new-joiner FAQ. Microsoft offers templates for this kind of assistant. Its My Company Policy agent template answers HR policy questions grounded in documents stored in SharePoint, with citations to the document and section, and its Employee Self-Service agent handles HR and IT questions. Our guide to Copilot Studio agents explains how such agents are built.

Two cautions. Microsoft's own Responsible AI FAQ for Employee Self-Service says it "isn't a substitute for decisions by HR, IT, legal, or other qualified personnel", and recommends a clear path to a person. And an agent repeats an outdated policy just as confidently as a current one, so keep the source documents up to date.

Learning plans

For learning and development teams, AI can turn a skills matrix into a learning plan per role, suggest a sequence, write short knowledge checks and draft session feedback forms. Keep the measurement honest: compare assessment results before and after training, not the number of courses completed. When the skill being taught is AI itself, teach judgement as well as features, for example with AI Fluency: Framework and Foundations.

A 30-day starting plan for an HR team

Week 1: set the rules. Approved tools, data that must never be pasted, who reviews AI output, and how to label AI-assisted documents.
Week 2: hiring pilot. Use AI for the job descriptions and interview kits of two open roles. Run the paired-resume bias test before any screening use.
Week 3: one policy. Rewrite a single policy in plain language and one Indian language, then send it for legal review against the Labour Codes.
Week 4: onboarding FAQ. Pilot a policy question assistant with a small group, log the questions it gets wrong, and decide whether to expand.

How CertTulen Academy can help

CertTulen Academy runs live, instructor-led AI training for HR and learning and development teams, online or as a private cohort at your office in major Indian cities. AI Fluency: Framework and Foundations builds sound judgement about what to delegate, the Prompt Engineering Masterclass covers structured prompts for job descriptions, rubrics and policy drafts, and Work Smarter with AI (AI-3025) gets the team productive in Copilot. See corporate training or contact us to plan a session.