AI Governance Should Enter the Room Before the Project Does

One of the things I’ve been thinking about as more companies move from AI experimentation to real investment is when governance should actually enter the conversation.

Too often, it comes in after the business case has been developed, the technology has been selected and the team is already thinking about implementation. At that point, governance becomes another review the project needs to get through.

I think that misses an opportunity.

If a company is considering an AI initiative, governance needs to be part of understanding the opportunity from the outset. That means looking at the data involved, the decisions the system will influence, the accountability around those decisions and the controls required to operate it responsibly at scale. These considerations can materially change how an organization evaluates the investment.

Those questions can materially change how attractive an initiative looks.

Click to expand full resolution

A use case with significant potential may also require substantial work around data, controls, security, operating processes or accountability before it can create value. Another opportunity may have a smaller headline benefit but a much clearer path to responsible deployment.

That information belongs in the investment decision.

This is why I see governance and prioritization as closely connected. Governance can help determine which initiatives are ready to move forward, which need preparation first, and which require a different approach altogether.

There is also a difference between building governance into the way an organization makes AI decisions and asking governance teams to review projects after the decisions have largely been made.

The first becomes part of the operating model. The second can easily become a checkpoint.

As AI becomes more capable and starts making or influencing decisions deeper inside the business, that distinction will matter even more. For me, good AI governance starts with a simple question i.e., What needs to be true for this AI investment to be worth making and responsible to operate?

That question belongs in the room at the beginning.