Let's start with a truth many association executives are already feeling, even if no one on staff has said it out loud yet: the board is going to ask about AI, and "we're looking into it" is not going to be a satisfying answer for much longer.

Maybe it already happened. A board member forwarded an article. A committee chair asked what the staff's AI policy is. A finance chair wants to know whether AI is a line item worth budgeting for or a risk worth insuring against. Whatever the trigger, the question underneath is the same one boards ask about anything material to the organization: who is accountable, and what does accountability actually look like here?

The conversation has moved. It's no longer about "Are we using AI?" It's about "How do we know we're using it well, and how would we know if we weren't?"

 Accountability Isn't a Policy Document  

Plenty of organizations have responded to board pressure by drafting an AI use policy. That's a good start, and I'd rather see one than not. The trouble comes when the policy becomes the finish line.

A policy is a speed limit sign. It tells people what's allowed. It doesn't tell you how fast anyone is actually driving, whether they're getting where they need to go, or who notices when something goes wrong.

Real accountability has three parts: a clear owner, a way to measure what's actually happening, and a cadence for reporting it upward. Miss any one of the three and your board is left holding a document instead of an answer.

Who Owns This?

I've sat across the table from enough leadership teams to know the most common failure here isn't negligence. It's diffusion. AI touches marketing, membership, education, finance, and IT all at once, so everyone assumes someone else is watching it closely. Nobody owns it, which in practice means nobody is accountable for it.

The instinct is to hand it to IT. It feels natural. AI is technology, IT owns technology, problem solved. I'd push back on that, and here's why.

Handing AI to IT and calling it governed is like handing the wiring to your facilities team and assuming they're accountable for everything anyone plugs in. Facilities should absolutely own the wiring. They should know the load the building can carry, which outlets are safe, and what happens when someone overloads a circuit. But they have no way of knowing whether the equipment plugged into the wall is producing anything useful.

That's the split worth drawing clearly.

IT owns the backbone. Which tools are approved, how they're licensed and secured, what data they're permitted to touch, how they connect to your AMS and the rest of your systems, and what a vendor's terms actually say about your member data. This is real, technical, non-negotiable work, and IT is exactly the right owner for it.

Business units own the work. Membership owns the renewal process. Education owns how course content gets developed. Marketing owns the campaign cycle. If AI is going to change how that work happens, the leader accountable for the outcome of that work is accountable for how AI shows up inside it. They are the only ones who know what "good" looks like in that process, and the only ones who will notice when a shortcut is quietly degrading quality. IT cannot see that from the outside, and shouldn't be asked to.

So the more useful question isn't "Who owns AI?" It's "Who owns the work that AI is changing?"

You still need one name the board can attach to the whole picture, whether that's a COO, a CIO, or a cross-functional group with a real chair and real authority. That person's job is to convene and hold the standard. Accountability that floats above the work, disconnected from the leaders who own the processes underneath, is just a lightning rod.

What Gets Measured

Here's where I'd steer most organizations away from their first instinct.

The easiest metrics to collect are adoption metrics. How many licenses. How many logins. How many staff completed the training. They're easy because they're already sitting in a dashboard somewhere. They're also close to meaningless on their own. Counting logins tells your board about as much as counting gym memberships tells you about anyone's fitness.

What the board actually wants to know is whether AI is creating efficiency in the organization, and whether you can show your work.

Pick two or three processes that matter and measure them honestly, before and after:

• Time. What did this process cost in staff hours before, and what does it cost now?
• Cycle time. How long did a member wait for an answer, and how long do they wait today?
• Capacity. What volume are you absorbing without adding headcount, and where did the recovered time go?

That last one is the one leaders skip, and it's the one boards care about most. Efficiency that disappears into a general sense that everyone is still busy isn't a result anyone can act on. Efficiency that shows up as a project you finally launched is.

Be just as honest in the other direction. If a tool created rework, drafts nobody could use, or output someone had to rebuild from scratch, measure that too. The report that includes what didn't work is the one your board will trust.

Imagine instead of a vague reassurance that "our team is exploring AI thoughtfully," your COO can tell the board that member inquiry response time went from three days to same-day, that first-draft time on course descriptions dropped by roughly half, that one pilot was shut down in March because it created more cleanup than it saved, and that the recovered capacity went into the recertification launch you'd deferred twice. That isn't a harder story to tell. It's just a story built on something real instead of good intentions.

Reporting Upward, on Purpose

The organizations that handle this well treat AI governance the way they treat financial reporting: a standing agenda item, a set cadence, a named owner, and a consistent one-page format the board can compare quarter over quarter. Comparability is the whole point. No board can spot a trend in a format that changes every time they see it.

The organizations that struggle treat it as a special presentation they scramble to assemble whenever a board member asks. The scramble itself sends a message, and it isn't the one you want to send.

A Question Worth Asking Before They Do

Sometimes the most useful question isn't "What is our board going to ask about AI?" but "What would we want to be able to say if they asked today?"

Try it at your next leadership meeting. Name the owner. Name the metric. Name the last time anyone reported on either one. If the honest answer makes you wince a little, that's not a failure. That's useful information, and it's far better to find it in your own conference room than in a board meeting.

Final Thought

Your board isn't asking about AI because they distrust your team. They're asking because AI has become material enough to the organization that oversight is simply part of good governance, the same as it is for finances or data security. That's a sign they're doing their job, not a sign you've fallen short at yours.

So give them something to hold onto. Not just a policy. A person accountable for it. Not just good intentions. A way to measure them. Not just an answer when asked. A story you're ready to tell before the question comes.