An example I made up
A director sits on the board of a regional consumer-finance company with its head office in Kuala Lumpur. It is the weekend, two days before the board meeting, and she is reading the board pack at home.
Near the back of the pack she finds an item marked "for noting". From next month, replies to customer complaints will be drafted by CitrilAI and sent without a second reader. Today, a second person reads every reply before it goes out. The paper quotes the tool:
"Sending replies without a second reader saves 900 staff hours a quarter and keeps reply quality at today's level. Based on last quarter's complaints and the quality team's checks. I am 87 percent sure."
The figure may well be right. What worries her is that the item changes who decides what a customer is told, and it asks the board to do nothing about it.
So she checks before the meeting. She sends the company secretary one question: who decided that this change belongs to management alone? The answer comes back the next morning. Nobody had decided it. The item was marked "for noting" because the system had already been bought. She asks one more thing: which complaints are in scope? The paper doesn't say whether complaints from customers in payment difficulty are included. She thinks those are the customers for whom a badly worded reply matters most.
Then she decides what to ask for. She writes to the chair. The letter asks for the item to move to "for decision", with the head of customer service named as its owner. It also asks for a monthly count of replies corrected after sending. She tells the chair and the chief executive before the meeting, so nobody sees the request for the first time at the meeting. At the meeting, the board approves the change for most complaints. It keeps a second reader for replies to customers in payment difficulty.
Three months later, the board has received that monthly count. Its agenda now marks as "for decision" every item where AI changes who decides something the board would want to decide.
Illustrative figures.
Why the label mattered
The trouble was the label. An item "for noting" tells the board that the right person has already decided. Once AI does part of the work behind a paper, it is worth asking whether the right person did decide. A tool can take a decision away from a person without anyone saying so. In the example, a second reader used to decide whether each reply was ready to send. The change removed the reader, and nobody took over the decision the reader used to make.
What I'd ask a board about its next meeting
These questions are about the board's own agenda and its own papers. They don't need anyone at the meeting to understand how the tool works.
I'd start with the agenda itself. Which items change who decides, because AI now does part of the work? Some papers say so plainly. Others don't. A cost saving or a system purchase can move a decision from a person to a tool, and the paper may describe it only as an efficiency.
For each of those items, I'd ask whether the board is deciding or only noting. "For noting" is a useful label. It saves the board's time on things that really have been decided by the right people. It shouldn't be used because a system has already been bought, or because a pilot went well. If an item changes who decides something the board would want to decide, it belongs under "for decision", with one person named as its owner.
Then I'd ask who outside the board has seen the answer the paper rests on. A board paper may now carry an AI answer, quoted or summarised. I'd also want to know whether the people who do the work have read it. In the example, the people who look after customers in payment difficulty could have said whether those customers were meant to be in scope.
Last, I'd ask what the board will look at in three months to know whether it decided well. If the board can't say what it will look at, it won't know later whether the decision was a good one. In the example, the answer was a monthly count of replies corrected after sending. It's a small thing to ask for. It tells the board what happened after the replies went out, which the paper does not show.
These questions ask the board to see where a decision has moved, and then to decide it at the meeting. AI can now prepare much of what reaches the board. The board still has to decide what belongs to the board, and say who owns the result.
AI can draft the plan. It cannot decide whether the plan is right.
The example above is made up, and I don't measure or score anyone.
I give keynotes, briefings and workshops on this.