Opinion
What should a board ask about AI-supported decisions?
Management says AI now supports its decisions. These are the questions a board can put to management to know whether those decisions are sound. None of them is technical.
Insights
I’m interested in what happens after AI gives you an answer.
I write about judgment at work: who takes responsibility for an AI answer, what a claim can support and why a proposal can fail.
My published pieces
Opinion
When is it right to automate a decision?
A senior leader is asked to let an AI decide a whole group of decisions alone. Here is the test I'd use on that proposal, and one made-up decision that shows why it matters.
Opinion
Why do reports on AI use hide how people decide?
A report on AI use counts how much AI is used. It doesn't show how anyone decided with what the AI gave them. Here's what that leaves out, and what a leader can ask for beside it.
Opinion
Sounding right is a style, not a signal
An AI answer now arrives finished. In my experience it sounds just as sure when it is wrong. That changes what a leader has to do before the answer becomes a decision.
Opinion
Why speed is not the same as progress with AI
An AI tool can compare three bids in an hour. The award still has to wait for the people who run the site. Here is one decision I made up, walked through, and what it says about speed.
Opinion
What a leader owes the people who must change
When a leader decides a change, the people who must carry it out are owed six things, and most of them are owed weeks before the change goes live.
Opinion
Why I stopped counting AI use as progress
A count of AI use tells you how often people opened the tool. It doesn't tell you what they decided with it. Here's an example I made up, and what I think a leader should decide about a target on use.
Opinion
What I would ask a board about AI before the next meeting
AI now does part of the work behind a board's papers. Before the next meeting, I'd ask the board which items change who decides, and whether it is deciding them or only noting them.
Opinion
What a manager owes a team that uses AI every day
When AI is part of every working day, the tool and the training are only the start. A manager owes the team five acts, and one of them is backing a check that slowed the work.
Opinion · Did not work
What I learned at twenty-nine when 241 per cent growth was not enough
At twenty-nine I delivered 241 per cent sales growth at Electrolux and was dismissed within the year. Here is what it taught me, and one made-up decision that uses the lesson now that AI drafts the plan.
Opinion
What I do when I am not sure the AI is right
Most AI answers reach me faster than I can check them. I still have to decide. Here's an example I made up, and what I do when I can't tell whether the AI is right.
Opinion · Did not work
What the Electrolux lesson means for AI projects
An AI project charter can be complete on its number and still leave out what the people whose work will change need. This is what my Electrolux story says to a leader before the work starts, while there is still time to decide what success means.
Opinion
What happens when a team stops checking
Checking an AI tool's work rarely stops because someone decides to stop it. It stops slowly, after many right answers, and other checks stop at the same time. Here is how that happens, in an example I made up, and what a leader can do before it costs something.
Opinion
What growing yearly recurring revenue from USD 4M to USD 25M+ in 12 months taught me
When a business grows fast, decisions arrive before the evidence for them. Decide now on the part that is urgent or easy to reverse, and give the part that can wait a date or a condition.
Opinion
The decisions I now make slower because of AI
AI answers now arrive in minutes, and they look complete. Most of the time that helps me. For three kinds of decision I go slower on purpose. Here are the three, and what slower means in practice.
Opinion
Your AI pilot worked. Three questions to ask before you scale it.
A pilot that worked is good news. It is not yet a reason to scale. Before I'd agree to a rollout, I want to know what the pilot was compared with, who checked the AI's work, and who owns the outcome when it goes wrong.
Opinion
What changing the work of 33,000 front-line employees taught me
A plan for front-line work can be checked from the staff's point of view alone. The customer lives with the same change differently, and that view needs a check too.
Opinion
What a good question to an AI sounds like
A one-sided question gets a one-sided answer, written well. A good question starts from the decision, asks for the case against, asks what the answer rests on and asks what would show it is wrong. Then someone still has to check and decide.
Opinion
What I ask when a number looks good
A planning assistant tells a sales director to copy one team's way of handling orders to two other countries. One question about how that team reached 99 per cent on-time delivery changes the decision. This is why I ask it, and the three questions I ask before I act on any number that looks good.
Opinion
What I tell young managers about being right too early
A correct finding still has to reach the person who decides, and who hears it first matters. I show this with one made-up decision, then say what I tell young managers.
Opinion
When the AI is faster than the team, who decides how fast the team works?
When AI output arrives faster than people can read it, the check becomes the slow step. A leader then has two decisions to make: where a person's check sits, and who owns what goes out.
Opinion · Did not work
What I would do differently at Electrolux
At twenty-nine I delivered 241 per cent sales growth at Electrolux and was dismissed within the year. Here is that year as one decision, and the check I would make now, before a good number is accepted.
Concession · Places
AI descriptions do not prove visitor spending
I explain why a description of a place cannot, on its own, show that AI changes what visitors spend.
Opinion
Why a leader should ask how people decided, not what they produced
A finished report looks the same whether or not anyone checked the work behind it. So the useful question for a leader is how the work was decided. Here is one example, and five plain questions to ask instead.
Opinion · Did not work
A price can be wrong even when the proposal is good
I look at why a proposal can fail on price, and why blaming the buyer is not enough.
Opinion
What scale does to a decision
A decision made once, for one customer, is a choice. Applied to every customer, the same decision becomes a rule. Here is what changes, shown through one example I made up.
Opinion
Who is responsible for an AI answer?
I argue that rules about tools and data do not settle who is responsible when someone acts on an AI answer.
Opinion
What 25 years across four regions taught me about decisions
An AI said one refund rule could serve five country offices, and its figures were right. One question to each office showed where the rule had to wait. This article is about that question.
Opinion
Free work still takes time
I argue that offering something free means considering the attention it will need after it is running.
I have 29 pieces here so far. The latest is dated 22 October 2026. I publish when I have something worth adding.
If one of these arguments matters to your audience, tell me about your event.