In many companies, employees already use AI assistants, often without saying so. One uses them for letters, another for spreadsheets, an intern to prepare presentations. Without shared rules, everyone does as they please, with real risks for confidentiality and the quality of work.
An AI use policy is a short document that sets the framework: what is allowed, what is forbidden and what must be checked. It does not need to be long to be useful. For an SME, two to three pages are often enough.
Why a policy rather than a ban
Banning AI completely is rarely effective: the use continues, but in secret, and the company loses all visibility. Conversely, allowing everything without a framework exposes it to data leaks and inaccurate documents. A policy lets you benefit from the time savings while limiting the risks. It also protects employees, who know clearly what is expected of them.
Take the example of a services firm in Bafoussam that handles payroll for several clients. Without a rule, an employee in a hurry might paste a payroll file into an online assistant to check the totals. With a clear policy, they know that this data is forbidden in public tools, that they must work on an anonymised template and that the final check is done in the payroll software. The rule does not stop them from using AI: it shows them how to do it safely.
The essential sections
- Purpose and scope: who the policy applies to, employees, interns, contractors.
- Approved tools: the list of assistants the company accepts, with the accounts to use, personal or business.
- Encouraged uses: writing drafts, rewording, summarising, help with spreadsheets, preparing training sessions.
- Forbidden data: customer information, staff data, detailed financial data, contracts, login details.
- Mandatory review: any content produced with AI is checked by its author before it is shared, and approved by a manager if it leaves the company.
- Decisions reserved for people: recruitment, sanctions, appraisals, financial commitments.
- Transparency: in which cases you must state that content was produced with the help of AI.
- The contact person and the procedure if something goes wrong: who to contact in case of doubt or an incident, for example sensitive data sent by mistake.
Writing the policy with the team
A policy imposed without explanation will be poorly followed. It is better to build it with the people concerned. Here is a simple approach:
- hold a short meeting to list current uses, without judgement;
- identify with the team the risky situations specific to your business;
- write a first draft, possibly using an AI assistant to help with the structure;
- have it reread by a few employees from different departments;
- present the final version with concrete examples.
Depending on the size and organisation of the company, it may be useful to have the policy checked by a legal adviser or an HR specialist, particularly regarding how it fits with the internal rules and personal data protection rules.
Keeping the policy alive
Tools change quickly, and a policy that never changes soon becomes outdated. Plan regular reviews, for example once a year or when a new tool is introduced. Include a presentation of the policy in the induction of new employees and interns. The contact person can also share good practices from time to time, such as an effective instruction or an error that was avoided.
Finally, think about training. A policy is not enough if employees do not know how to anonymise a document or check an answer. A short hands-on session of an hour or two can make a big difference.
An AI use policy is not just another administrative document. It is a tool for trust between management and the team, which lets everyone use these technologies in a useful and responsible way, knowing exactly where the limits are.