AI Governance

AI Governance

Artificial intelligence is already making decisions inside your organization. The question is no longer whether you adopt it — it is who answers for its decisions.

Every week, AI systems approve, rank, recommend and execute across businesses: credit decisions, pricing, candidate screening, customer responses. When one of those decisions causes harm, the chain of accountability rarely leads to a name. The model belongs to the vendor, the algorithm to the technical team, the policy to compliance — and the decision itself belongs to no one.

That is the gap AI governance has to close, before an incident, an auditor or a regulator closes it for you. (Further reading: “The Decision Nobody Owns,” published on this site.)

Who this is for

  • Boards of directors that must exercise real — and demonstrable — oversight of the organization’s AI systems, without becoming technical committees.
  • Executive teams deploying AI who want to know, decision by decision, who is accountable for it.
  • Financial institutions and insurers, for whom the regulatory framework is already taking shape: OSFI’s Guideline E-23 on model risk management takes effect on May 1, 2027, and Quebec’s AMF is moving in the same direction.
  • Structured SMEs and growing organizations adopting AI tools faster than their control mechanisms.

What the mandate includes

1. AI governance diagnostic
An inventory of the AI systems in use — including the ones nobody officially approved — a map of the decisions they make or influence, and a gap assessment against regulatory expectations and defensible practice.

2. Decision rights
For each category of automated decision: who may delegate it to a system, under what conditions, with what level of human oversight, and who can stop it. Written rights, assigned to named individuals — not to vague functions.

3. Accountability framework
A reporting structure that holds up before a board, an auditor and a regulator: decision registers, escalation thresholds, review and challenge mechanisms, and the documentation that proves oversight actually took place.

4. Board and executive support
Implementing the framework with the people who will carry it: working sessions with the board or the relevant committee, preparing executives for auditor and regulator questions, and adjustments as the systems evolve.

A successful engagement leaves you able to carry on without us. We transfer methods, tracking tools and decision habits to your teams, and the end of the engagement is planned from the day it starts. Everything we build together is yours to keep: analysis grids, policies, dashboards and documentation.

The approach: Act, Don’t Be Acted Upon

« Agir, pas subir » — Act, Don’t Be Acted Upon. Organizations that are acted upon by AI let it in through the tools door, without ever deciding how it will decide once inside. Those that act set the rules first: which decisions get automated, which never do, and the name of the person who answers for both.

We approach every engagement as people who have run organizations. Before recommending anything, we check what is achievable with your teams, your systems and your timeline, and we weigh what each option costs in time, attention and risk. Our recommendations are built to be applied quickly, and we stay involved until they are in place.

First conversation

Every mandate begins with a first, confidential, no-commitment conversation: where your AI systems stand, who answers for them today, and what a defensible framework would concretely require in your organization.

Jean-Philippe St-Jacques — 438 801-2614 — jps@5mindset.com