Governance That Enables Innovation
Organizations are moving quickly with AI, and governance has to keep up. I am interested in how organizations make good decisions about where AI should be used, what level of oversight is appropriate, who is accountable, and what needs to happen as systems change over time.
Good governance should give people enough clarity to move forward. When responsibilities, decision rights and expectations are understood, teams can spend less time figuring out the process and more time making sound decisions about the technology.
That is the kind of governance I am interested in building.
Current Areas of Focus
My interest in AI governance comes from years of working with technology inside large organizations. Policies matter, but eventually someone has to make a decision, take responsibility for it and deal with what happens after the technology is put into use. These are some of the areas I am looking at more closely.
Who Makes the Decision
AI decisions need clear ownership. Who can approve a use case? Who can ask for more evidence, place limits on its use or decide that it should not move forward? I am focused on what happens when technology, business and risk leaders see the same decision differently.
Matching Governance to Risk
I don't think every use of AI needs the same level of review. An internal productivity tool is very different from a system that can influence a financial, employment or clinical decision. The governance should reflect what the system does, who it affects and what could happen if it gets something wrong.
After Approval
An approval is a decision made at a particular point in time. The model may change. The data may change. The way people use it may change. I am interested in how organizations decide when those changes are significant enough to require another review.
Human Oversight
I am skeptical of treating “human in the loop” as an answer by itself. Meaningful oversight depends on whether the person actually understands what the system is doing and has the authority, information and time to challenge it. This becomes especially important in healthcare, where AI can begin to influence clinical judgment and the way clinicians work.
When AI Fails
Organizations also need to know what happens when an AI system produces an unexpected result or creates a problem. Who investigates it? Who decides whether the system can remain in use? Who communicates the issue, and who has the authority to restrict or stop it? Those responsibilities are much easier to establish before an incident occurs.