Mohammad Manzoorul IslamFounder & Chief Consultant, Mahdiya · November 2026
Ask any team in Dhaka whether they use AI at work, and most will say no. Ask them how they wrote last week’s client email, summarised a long report or translated a document, and the answer changes. ChatGPT, Gemini, Copilot and Claude are already in daily use, often on personal accounts, with no rules about what may be pasted into them.
This is called shadow AI, and it carries three real risks:
- Data leaving the company. Customer lists, contracts, salary sheets and source code pasted into a free AI tool may be stored outside your control.
- Wrong answers sent as facts. AI tools write confidently, including when they are wrong. A figure or clause that nobody checked can reach a client or a regulator.
- Client and regulatory exposure. Banks regulated by Bangladesh Bank, BPOs with EU clients and exporters under foreign contracts are increasingly asked how they control AI use. “We have no policy” is a weak answer.
Banning AI rarely works. People keep using it, just less openly. A short, clear policy works better: it says what is allowed, what is not, and who to ask.
What a good first policy does
A first policy does not need to be long. It needs to answer five questions every employee has:
- Which AI tools may I use for work?
- What information must never go into them?
- When must a person check the output before it is used?
- When do we tell clients or customers that AI was used?
- Who do I ask, and how do I report a mistake?
The template below answers all five on one page. It is a starting point for small and mid-sized companies, not legal advice. Adjust it to your sector, your contracts and any rules your regulator sets.
The template
[Company name] AI Acceptable-Use Policy Version 1.0 · Effective [date] · Owner: [name, role] · Review: every 6 months 1. Purpose We use AI tools to work faster and better. This policy keeps our clients' data, our staff and our reputation safe while we do. 2. Approved tools You may use only these AI tools for work: [list, e.g. ChatGPT Team, Microsoft Copilot]. Use your company account, never a personal one. Ask [AI contact] before using any other tool. 3. Never enter into an AI tool - Customer or employee personal data (names with phone numbers, NID, account or card numbers) - Passwords, system access details or source code - Contracts, pricing, financial results or board papers that are not public - Anything a client contract or regulator says must stay confidential 4. Check before you use A person must read and check AI output before it is sent outside the company or used for a decision. You are responsible for anything you send, whether AI drafted it or not. 5. Be open about AI Tell clients or customers when they are talking to an AI system, or when content they receive was mostly generated by AI, where a contract, a law or common sense requires it. 6. Not allowed Using AI to make final decisions about hiring, credit, discipline or pricing without human review; creating fake images, voices or documents of real people; or getting around this policy with personal tools. 7. Ask and report Questions: [AI contact, email]. If data was entered by mistake or AI output caused a problem, report it to [AI contact] within 24 hours. Honest reporting will not be punished. Approved by: ____________ (CEO) · Date: ________
The template stays on one page so staff actually read it. Sector-specific rules (for example Bangladesh Bank requirements for banks and MFS providers, or EU AI Act duties for BPOs serving EU customers) go in an annex, not in the main page.
How to put it in place in two weeks
- Week 1: List the AI tools your teams already use (a short survey is enough). Choose the approved ones, ideally business accounts where your data is not used for training. Fill in the blanks in the template.
- Week 2: Have the CEO sign it, share it with all staff, and run a 30-minute briefing. Name one person as the AI contact. Review the policy after six months.
This policy is the first step toward a fuller AI management system such as ISO/IEC 42001, which adds risk assessment, monitoring and continual improvement.
Want help adapting it?
Mahdiya helps organisations adapt this policy, build an AI tool inventory and assess their AI readiness.
Talk to Manzoor