AI rarely becomes a board problem overnight. It builds quietly for a couple of quarters and then one event, a bill, an incident, an audit question, drags it into the room.
By the time it lands on the agenda it is usually expensive and a little embarrassing to explain. The good news is that it gives you plenty of warning first. I see the same signals over and over, and honestly most leadership teams have at least two of them right now without quite naming them. Here are the five I watch for.
1. Nobody actually owns the AI spend
Ask who owns the AI line in the budget and watch what happens. If you get a pause, or three different names, or a vague wave toward IT, that is the sign. The spend is real and it is growing, copilots, agents, vendor subscriptions, tokens, all of it, but it is scattered across cards and cost centres and nobody is accountable for the total. The reality is you cannot manage a number you cannot see, and the board will eventually ask for that number.
2. The CEO and the CIO are scared of opposite things
This one is almost funny when you spot it in a room. The CEO is frustrated that AI is moving too slowly and the competitors are making announcements. The CIO is awake at night about data leakage and tools they cannot see. Both fears are completely valid and they pull in opposite directions, so nothing decisive happens. That standoff is its own warning, because it usually means there is no shared picture both of them can actually trust.
3. You genuinely cannot list the tools in use
If I asked you to write down every AI tool running across the business today, with what data each one touches, could you do it. Most people cannot, and they know it the moment the question is asked. People are sensible and they are also under deadline pressure, so they reach for whatever helps, approved or not. That is not recklessness, it is just convenience winning. But every tool you cannot see is a little bit of risk you are carrying blind.
4. Lots of pilots, almost nothing in production
Plenty of firms can point to a wall of pilots and demos. Far fewer can point to AI that is actually running in production, owned, measured, and relied on. If your pilots keep stalling just before the finish line, something structural is missing between the experiment and real capability, and it is usually ownership and a path to production rather than the technology. A pile of stalled pilots is money spent with no return, and someone on the board will eventually notice.
5. One or two people are quietly carrying all the value
Look at where the real AI wins are coming from. Very often it is one or two people, and they are not always senior. An analyst, a developer, someone in ops who figured out how to create enormous leverage. That is great until you realise how exposed you are. If you clamp down on them with heavy rules they get frustrated and leave, or they go quiet and hide the good work. Either way the value walks out the door, and losing your best builder is a far bigger problem than most boards realise until it happens.
What to do before it lands on the agenda
None of these signs mean anything has gone wrong yet. They mean you still have time, which is the whole point of noticing them early. Early visibility is cheap. Late visibility, the kind that arrives with an incident or a surprise invoice attached, is expensive and it comes with an awkward board conversation on top.
The first move is just to see it clearly. Map where AI is actually being used, what it is costing, where the risk is gathering, and who is creating the value, then put a plain picture in front of leadership before someone else puts a nasty surprise there instead. That is exactly what a FitCheck is built to do, and it is a few weeks of work, not a transformation program.
If two or three of these felt a bit close to home, that is worth a conversation. Book a working session and we will work out where you actually stand.