AI at work

Company AI Usage Policy: A One-Page Starter Template for Indian Firms

By Sauravjit Singh · Updated 2026-11-04 · 7 min read

Short answer: A company AI usage policy tells employees which AI tools they may use, what data they can and cannot put into them, who checks AI output before it reaches clients, and who to ask when unsure. For a 20–50 person firm, one plain-English page plus a green, amber and red data table is enough to start, reviewed by a lawyer.

Most small companies already have staff using ChatGPT, Gemini or Copilot on personal accounts, often with client files. Banning AI rarely stops that; it just pushes it out of sight. A short policy that says yes, with clear limits, is easier to follow and easier to enforce.

Below is a starter policy you can adapt, the data table that does most of the work, and pointers on India's data protection law. This is a practical template from an operations point of view, not legal advice.

What should a company AI policy cover?

Eight things, and each can be a sentence or two. Longer policies tend not to get read.

  • Purpose: why the company allows AI and what it wants from it.
  • Approved tools and accounts: which tools, and whether company accounts are required.
  • Data rules: what may go in, what needs anonymising, what never goes in.
  • Human review: who checks AI output before it reaches a client, candidate or regulator.
  • Disclosure: when to tell clients or colleagues that AI helped.
  • Ownership and confidentiality: work made with AI is company work, under the usual confidentiality terms.
  • Owner and questions: a named role who answers questions and approves new tools.
  • Breaches and review: how to report a mistake, and when the policy is reviewed.

The one-page starter policy (adapt the wording)

This version suits a 20–50 person services firm, say a Noida marketing agency or a Pune accounting practice. Replace the role titles with yours and remove anything that does not apply.

  1. Why we use AI. We use AI tools to save time on drafting, research, summarising and analysis, so people can spend more time on judgement and client work. AI assists; people decide.
  2. Approved tools. Use only the tools on the approved list kept by the Operations Head. Use company accounts for work. Personal accounts are not allowed for any amber or red data.
  3. Data. Follow the green, amber and red table. If you are unsure which colour something is, treat it as red and ask.
  4. Check everything. You are responsible for any output you use. Check facts, numbers, names, legal or tax points and links before anything leaves the company.
  5. Professional advice. AI output on tax, GST, legal or HR-compliance matters is a draft only. It must be reviewed by the responsible qualified person (CA, lawyer or HR lead) or checked against the official source before use.
  6. Clients. Do not upload a client's documents into any AI tool unless the client agreement allows it or the client has agreed in writing.
  7. Disclosure. Tell your manager when AI did a substantial part of a deliverable. Follow any client's own rules on AI use.
  8. No AI for decisions about people. Hiring, appraisal ratings, pay and discipline decisions are made by people. AI may help draft text from human-supplied evidence.
  9. Mistakes. If you put red data into an AI tool by mistake, tell the Operations Head the same day. Quick reporting is not punished; hiding it is.
  10. Review. This policy is reviewed every six months, because tools and laws change.

Get each employee to acknowledge it, and add it to the induction pack for new joiners. In our experience running teams since 2013, a policy no one is walked through is a policy no one follows.

The green, amber and red data table

This table answers the question employees actually ask: can I paste this? Print it and pin it.

ColourExamplesRule
Green: public or harmlessPublished website copy, public news, generic templates, your own rough ideas, made-up sample dataFine in any approved tool
Amber: internal or identifiableInternal process documents, anonymised client data, draft proposals, sales figures without names, anonymised resumesApproved company accounts only, with model training switched off; remove names and identifiers first
Red: personal, confidential or regulatedCustomer or employee personal data (phone, Aadhaar, PAN, bank details, health), client financials, passwords and API keys, unreleased contracts, source code under NDA, anything a client said is confidentialNever in a general AI tool, unless the Operations Head has approved a specific tool and contract for that use

A common gap: people forget that screenshots, spreadsheets and email threads contain red data too. The rule applies to uploads, not just typed text.

Which AI tools and accounts should you approve?

Approve one or two general assistants on business plans rather than a long list. As of November 2026, the major assistants (ChatGPT, Claude, Gemini, Copilot) offer business or team tiers whose terms generally say your inputs are not used to train their models by default, plus admin controls. Terms change, so read the current data and privacy terms for each tool before approving it, and record the date you checked.

Things to confirm for each tool: whether inputs are used for training, how long chats are stored, whether admins can remove users who leave, and where data is processed if your clients care about that. If you use Google Workspace or Microsoft 365 already, check what AI features your existing plan includes before buying something new.

What does the DPDP Act mean for AI use at work?

In short, if employees put personal data of customers, candidates or staff into AI tools, your company is responsible for that processing under the Digital Personal Data Protection Act, 2023. The Rules under the Act were notified in 2025 with obligations phasing in, so check the current status on the MeitY website and confirm with a lawyer. Pointers worth raising with counsel:

  • Your company is likely a 'data fiduciary', the entity that decides why and how personal data is processed. An AI vendor you use is typically a processor acting for you.
  • Use personal data only for the purpose it was collected for, with notice to the people concerned.
  • Put reasonable security safeguards in place, which includes controlling which tools staff use.
  • Have a process for reporting personal data breaches, which under the Act can include notifying the Data Protection Board and affected people.
  • Regulated sectors (lending, insurance, healthcare, listed companies) may have additional rules from their regulators.
  • Client contracts often restrict where their data can go; these can be stricter than the law.

This list is a starting point for a conversation with your lawyer, not a compliance checklist.

How do you roll the policy out without slowing people down?

  1. Ask the team, anonymously, which AI tools they already use and for what. You will learn where the real risk sits.
  2. Approve tools that cover those uses, so people have a permitted option.
  3. Run a 30–45 minute session walking through the data table with examples from your own work.
  4. Collect a signed or emailed acknowledgement from each person.
  5. Give a short set of approved prompts for common tasks, so the safe way is also the easy way.
  6. Review after six months: what tools were added, what mistakes were reported, what needs clarifying.

The training step matters more than the document. Our guide on how to train employees on AI covers a practical format, and the employee side of the same question is in our piece on whether to tell your boss you use AI.

Mistakes founders and HR teams make with AI policies

  • A blanket ban. It drives use onto personal phones and accounts, where you have no control.
  • Copying a 12-page enterprise template. Nobody in a 30-person firm reads it.
  • No named owner. Questions go unanswered and people guess.
  • No examples. 'Do not share confidential data' means different things to different people; the table fixes that.
  • Forgetting HR uses. Resume screening and appraisals involve personal data; see our guide to AI in HR work.
  • Never reviewing it. Tools add features every few months.

For the wider picture of where AI fits across departments, see our guide to how to use AI at work. If you would rather have staff trained on safe AI use by an outside team, our corporate AI training page lists the workshop formats available.

Frequently asked questions

Do Indian companies need an AI policy by law?

As of November 2026 there is no general Indian law requiring a standalone AI policy, but data protection, sector rules and client contracts all apply to how staff use AI. A written policy is the simplest way to show you have reasonable safeguards. Confirm your obligations with a lawyer.

Can employees use free ChatGPT for work?

For green, public data it is usually low risk. For anything internal or personal, a company business account with training switched off is safer, and your policy should say which data is allowed where.

Is there a free AI policy template for small companies?

The one-page version in this article is free to adapt. Have a lawyer review it against your client contracts and sector rules before you adopt it.

What happens if an employee pastes client data into ChatGPT?

Treat it as an incident: record what was shared, delete the chat if possible, check the tool's data terms and assess with your lawyer whether the client or authorities need to be told. A no-blame reporting rule makes people report quickly.

How often should an AI usage policy be updated?

Every six months is a practical rhythm for a small firm, plus whenever you approve a new tool or the law changes.

Want to learn this with a trainer?

Leave your details and our counsellor will call you within one working day with batch dates, fees and the right course for you.

  • Free counselling call, no obligation to join
  • Weekday and weekend batches, online or offline
  • Small batches of 15 to 20 students

Or chat on WhatsApp

Enter a 10-digit mobile number.

We use your details only to contact you about Spettro School courses.

Chat on WhatsApp