top of page

AI Governance: Where to Start? (A Guide for Executive Management and Boards of Directors in Quebec)

Photo by the editor: Sarah Mjidou
Sarah Mjidou
Oct. 1
5-minute read

Last updated: Oct. 6

AI is already present in your organization, with or without guidelines. The real question is: where do you start to manage it? Here’s a four-step approach to building an AI governance framework that accelerates rather than hinders progress—and the role your board of directors plays in it.


In most organizations in Quebec, artificial intelligence has made its way in without waiting for permission. Employees are drafting their emails using a generative AI tool. Managers are summarizing reports with a chatbot. And many software programs you’ve been using for years now include AI features, which are often enabled by default.


Meanwhile, the board of directors is starting to ask questions. What tools are we using? What data is flowing through them? Are we in compliance with Act 25? Management doesn't always have a clear answer.


Two approaches tend to dominate in such situations: banning everything or letting things run their course. Banning drives usage underground. Letting things run their course exposes the organization to risks it cannot fully assess. AI governance is the third way: when well-designed, it is not a hindrance; rather, it is what enables us to move forward with confidence.


What AI Governance Is (and Isn't)


AI governance refers to the set of rules, roles, and mechanisms that govern how your organization selects, uses, and monitors artificial intelligence. It addresses four questions: Who makes the decisions? What is permitted? How are risks assessed? And how do we ensure accountability?

It’s not a 40-page document that no one reads. Nor is it a matter to be left solely to the IT department. AI affects human resources, legal affairs, operations, and customer relations: its governance must be cross-functional and led by senior management.


The 4 Steps to Building Your AI Governance Framework


1. Take stock of actual practices: You can’t govern what you can’t see. Before writing anything, paint an honest picture of the situation:

• Which AI tools are used, whether officially or unofficially;

• what data is entered there: personal information, confidential information, financial data;

• What AI projects are currently underway or being considered;

• Which vendors have added AI features to your existing software.

A short, anonymous survey of employees often provides a more accurate picture than the official assessment. The goal is not to assign blame, but to understand needs. If your teams are using AI on the sly, it’s because they see value in it that the organization isn’t yet providing.


2. Adopt a clear AI usage policy: This is often the first concrete deliverable—and the one with the most immediate impact. To ensure it is read and followed, it should be no more than a few pages long and written in language that employees can understand. In particular, it should specify:

•         authorized tools, and the process for getting new ones approved;

•         Prohibited Data : personal information, confidential information about clients or partners, trade secrets;

•         human verification : all AI-generated content is reviewed and approved by a person who takes responsibility for it;

•         Transparency : cases where it is necessary to indicate that content or a decision is based on AI;

•         Reporting Procedures : Who to contact in case of questions or incidents.

 

3. Incorporate Bill 25 from the design phase: In Quebec, AI governance is inextricably linked to the protection of personal information. Bill 25 does not specifically target AI, but many of its requirements directly affect AI projects.

The practical lesson: Incorporate these checks into the AI project approval process, rather than conducting them after the fact.


4. Define roles and manage risks in a proportionate manner: Effective governance relies on clear responsibilities. In most organizations, this involves a small AI committee that brings together business, IT, information security, legal, and human resources, under the leadership of an accountable executive.

This committee must not become a bottleneck. The key is proportionality: not all uses pose the same level of risk.


Risk Level

Examples of usage

Suggested Requirements

Low

Rewrite an internal memo

Compliance with the Terms of Use

Medium

Draft customer communications, analyze internal data

Manager approval, systematic human verification

High

Screening applications, evaluating a client's file—any decision that affects a person and has an impact on operations

In-depth analysis, enhanced human oversight


Keep a record of approved use cases. Also, integrate AI into your procurement processes: each supplier should be able to explain where your data is hosted, whether it is used to train its models, and how to disable its AI features.


Success Story: Quasi-Governmental Organization


Our mandate: an AI and data governance framework aligned with Bill 25.

What this illustrates: AI governance and data governance go hand in hand. An organization that doesn’t know what data it holds, where it is located, and who has access to it will struggle to regulate AI in a credible manner. Conversely, well-governed data accelerates all subsequent AI projects.


The Board of Directors and AI: Monitor Without Managing


The board of directors is not responsible for managing AI, but it must ensure that it is properly governed. As with cybersecurity, this is a matter of risk oversight and strategic direction.

Two simple steps can make a big difference. First, put AI on the agenda on a regular basis and specify which board committee is responsible for overseeing it. Second, schedule a training session for board members: it’s hard to oversee a technology you don’t understand.


⚠️ Five Questions to Ask at the Next Board Meeting


• Do we know where and how AI is used within the organization?

• Who in management is responsible for this?

• Is our risk appetite for AI defined and documented?

• How would the board be notified of an AI-related incident?

• Is AI part of our strategy, or is it treated as merely a technological issue?


Frequently Asked Questions


Does an SME really need AI governance?


Yes, but on a scale appropriate to the company. An SME does not need a formal committee or a comprehensive framework. A clear usage policy, an inventory of tools, and a designated person in charge already provide a solid foundation, especially if the company handles personal information.


Should consumer AI tools like ChatGPT be banned entirely?


A total ban is rarely effective: usage simply shifts to personal devices. It’s better to offer an approved solution and clearly specify what can and cannot be entered into it.


How long does it take to set up an initial framework?


It depends on the size and complexity of the organization. A first functional version (inventory, policies, roles) can generally be implemented within a few weeks. Maturity, however, is built up over several months, as projects progress.


Start small, but start now


AI governance is not a compliance exercise that can be checked off once and for all. It is a capability that enables organizations to capitalize on the opportunities presented by AI without losing the trust of employees, customers, and the board.


Wondering where to start in your organization? Schedule an initial 30-minute conversation with Alnixia. We’ll assess your situation and discuss the most helpful next steps.

bottom of page