AI reads a resume against an opening, explains a payroll flag, reads a receipt or a tax proof and drafts a helpdesk reply, on the screen where the work happens. The assistant answers from your company's data within each person's role. Every result is a suggestion, and a person decides.
A sample chat. An employee asks for a day of casual leave, checks the request and confirms it. A payroll admin asks why payroll cannot be released and is told who must approve it.
Each one sits in its module and is switched separately from the chat. These are sample screens: the labels and states are the product's, and the names and figures are made up.
A score and reasons on the candidate page, against the opening's requirements: what the resume shows and what it does not.
Plain words for the checks that fired on a computed run, and what to check first.
Reads the amount, date, vendor and category from an uploaded bill, and warns if it may already be on another claim.
Suggests a category when a ticket is raised, and drafts a reply from your knowledge base when an agent asks.
Type what you want and AI picks the report or sets up the custom report. It returns a report definition, not data.
Reads the section, amount, financial year and issuer from an uploaded proof, and shows how they compare with what was declared.
Feedback on each goal for a manager writing a review, goal ideas, a 1:1 summary, and wording for a letter template. Letter drafts can be in English or Hindi, though issued letter PDFs cannot print Hindi yet.
The themes in a closed survey's open-text answers, and actions HR could take. It shows no names and no quotes.
Plain words on each department's exit figures, and what to check. It explains past trends. It is not a prediction, and it says nothing about any person.
Courses from your own catalogue for each person, from their role and skills.
Suggests which NeevHR field each column of an employee or migration file holds, and fills a draft leave policy from your policy PDF.
NeevHR switches a module's AI on for your company, and your AI admin can turn it off. With it off, the screen is the normal complete screen, with no empty box and no prompt.
It never rejects, approves, pays or changes a record by itself. Every result is labelled as an AI suggestion, with the date it was made.
The same record with the same input never calls the model twice. The stored result is reused at no charge, and nothing calls the model when a page opens.
AI text shows only to someone who may already open the record and the fields it was built from. Disciplinary and POSH have no AI.
AI runs on credits your company buys. When they finish, stored results still show, no new run starts, and the feature keeps working without AI.
When your company switches a feature on, this goes to our AI provider, Anthropic, which may process it outside India. Files are sent whole, and a file can contain what is printed on it, so ask people to upload only what the claim or proof needs. See the privacy policy.
| Feature | What is sent |
|---|---|
| AI assistant (chat) | The question, and the record data needed to answer it within that person's role, such as their own payslip summary or leave balance. The assistant's tools do not read bank account, IFSC, PAN, Aadhaar, UAN, ESI number, address, tax details, health or disciplinary data, a candidate's phone or email, or another person's individual pay. Whatever a person types in a question goes as typed. |
| Resume match | Cleaned resume text and the opening's title and job description. Automatic rules take out the name, contact details, ID numbers, address, personal details and pay first. Rules work on patterns, so something written in an unusual way could remain. |
| Expense receipt reading | The whole receipt file as uploaded (image or PDF), and your expense category names. No detail is added from the record, but a file can contain what is printed on it, such as a name or a card or account number. |
| Tax proof reading | The whole proof file as uploaded (image or PDF), and the product's tax section names. A file can contain what is printed on it, such as a name, PAN or account number; those are not kept or shown from the reading. Proofs on the health-related lines (80D, 80DD, 80DDB, 80U) are not sent. The check is by the line a file is uploaded on, so a health paper uploaded on another line would go. |
| Leave policy from a PDF | The whole PDF as uploaded, and the names of your leave types. A PDF whose text shows a PAN or Aadhaar number is refused. A scanned page has no text to check, so it can still go if it shows one. |
| Import column matching | Column headings, and up to three masked samples per column that show the shape of a value, not the value. Columns for PAN, Aadhaar, bank, pay and other personal details send no sample. The rows of the file do not go. |
| Payroll change explanation | The checks that fired, with a line reference for each person, the percent change in net pay, loss-of-pay days, overtime hours and variable pay as a percent of gross. No name, employee code, department, bank detail, PAN or rupee amount. |
| Helpdesk category and reply draft | The ticket subject and the conversation with the employee, after automatic rules remove names, contact details and ID numbers, your ticket category names, and matching knowledge base and Help Centre text. Internal notes and attachments do not go. Confidential categories and tickets that read as a private matter do not use AI. |
| Report from a sentence | The sentence typed, and the names of the reports and fields the person's role may use. No employee data. A sentence with a person's full name, code, email, phone or ID number is refused. |
| Review, goal and 1:1 drafts | For a review: each goal's title, description and weight, and the employee's own note and self percent. For a goal: the designation, department and existing goal titles. For a 1:1: the agenda, shared notes, action items and that person's own private note. Automatic rules remove names, codes, contact details, ID numbers, bank details and pay figures first. Ratings, pay and the other person's private note do not go. |
| Letter template draft | The letter type, language, template name and category, and the names of the merge fields. No employee data: details stay as merge fields until a letter is issued. |
| Survey summary | The question text and open-text answers, with no name, employee code or department attached, after automatic rules remove names, contact details, ID numbers and pay figures from the words. A question goes only with at least the survey's minimum group of comments. An answer that reads as a health, private, disciplinary or POSH matter is left out. |
| Attrition reasons | For each department large enough to show: its name, average headcount, exits, attrition percent, trend, and counts of exit types and resignation reasons. No name, employee code, pay, rating or exit interview text. |
| Learning suggestions | The person's designation, department and skill names, and your catalogue's course titles, types, hours and skills. No name, employee code, contact detail, pay, rating or attendance. |
Employees stop waiting for HR on routine questions, and anything that would change a record comes back to a person to confirm. Your admin decides the rest.
It is switched on module by module, and your AI admin can turn a module off. Your Company Admin has it from day one and decides who else gets the assistant: everyone, or only the people you name.
AI runs on prepaid credits your company buys, never the employee. You see the balance at any time, and when the credits finish, AI stops. There is no surprise bill.
A usage report by person and department, with date and module filters, ready to export. It shows questions asked and credits used, never what anyone asked.
Exactly what the person's role can open on screen. The assistant's tools do not read another person's pay, or bank, PAN, Aadhaar, UAN, ESI number, address, health and disciplinary data. An uploaded receipt or proof is sent whole, and a file can contain what is printed on it. Anyone who tries a trick to get more is warned once, then blocked for a day.
Resume match, payroll change explanation, receipt and tax proof reading, helpdesk drafts, report from a sentence, review and letter drafts, survey summary, attrition reasons, learning suggestions and setup from Excel or PDF. See them below.
Leave balance, attendance, shift, holidays, expense claims, tickets, courses, and their own payslip summary, CTC, loans, goals and released rating.
Why payroll cannot be released and who must approve it, who a request is waiting on, why someone cannot see a record, and whether an approver has been set.
It prepares a leave request, a regularization or a helpdesk ticket as a review card. Managers review leave decisions by name. Nothing happens until the person confirms.
How-to and policy answers from the Help Centre and your company's documents, with the article cited. When it is unsure it says so and points to HR.
It replies in the language the person writes in.
In Payroll, Compensation, Variable pay, Benefits, Loans, Recruitment and Performance it only reads. It cannot run payroll, change a salary, make an offer or change a rating.
NeevHR switches it on for your company, module by module, and your AI admin can turn a module off. Your Company Admin has it from the start.
Your admin decides who may use it. Each person reads a notice on what is sent before first use.
People ask in plain words, or see a suggestion on the record they are working on, with the same permissions they have on screen.
Anything that would change a record comes back as a review card, and only the person's confirmation carries it out.
Set as rule sets by your HR admin, assigned to employee groups.
Every list becomes a report; every report a chart that drills to records.
Like every module, it sits on one effective-dated employee record, with India's statutory rules, configuration and governance shared across the platform.
PF, ESI, PT, LWF and TDS, gratuity, bonus and the DPDP Act 2023 are part of the core product, not an add-on. INR and the India financial year throughout.
Process discipline and statutory accuracy for growing and mid-market companies, without the cost or the multi-year rollout of a tier-one suite.
Implementation time depends on employee count, data quality, integrations and payroll complexity.
Leave, attendance, payroll, expense and appraisal are configurable rule sets assigned to employee groups. A trained HR admin can own them, no certified consultant required.
No. It never gives another person's individual pay, for any role, whoever asks. Company pay appears only as totals, and small groups are hidden so nobody's pay can be worked out.
No. Each check follows the permission of the screen behind it. Only people who manage users or roles can check another person's access. Anyone else is told to ask their HR admin.
The provider is Anthropic, which may process data outside India. For the assistant: the question and the record data needed to answer it; its tools do not read bank account, IFSC, PAN, Aadhaar, UAN, ESI number, address, health or disciplinary data. For a module's AI, when your company switches it on: cleaned resume text, and uploaded receipts or proofs as whole files. A file can contain what is printed on it, so it can show a name, PAN or account number. Proofs on the health-related lines (80D, 80DD, 80DDB, 80U) are not sent. The table on this page lists each feature.
No. It never rejects, approves, pays or changes a record. Each result is labelled as an AI suggestion, and a person uses it or ignores it.
No. Nothing calls the model when a page opens, and the same record with the same input reuses the stored result at no charge. Some features run by themselves once switched on, such as resume match when an application arrives; these use credits and have a monthly limit per feature.
No. NeevHR switches a module's AI on for your company, and your AI admin can turn it off. With it off, the screen is the normal complete screen, with nothing about AI.
No. The usage report shows how many questions a person asked and the credits used, never the questions or the answers.
Your company buys prepaid credits. Each answer uses credits according to its length, and only the company recharges, never the employee. See the pricing page.
One platform, one employee record.
A tailored walkthrough for your team, no generic deck.