Judgement: The skill AI can't replace
Free masterclass page 1 of 3
Every day, you make calls that no algorithm can make for you.
Not because the technology isn't there. It is. The models are extraordinary. They can process a thousand data points before you finish reading this sentence. They can draft your strategy document, surface patterns in your customer data, and generate three plausible scenarios for your Q3 forecast before lunch.
But they cannot tell you which scenario to believe. They cannot read the room when your CFO's hesitation means something the spreadsheet doesn't show. They cannot feel the weight of a trade-off between two values that both matter and pull in opposite directions. And they cannot decide, on your behalf, when the elegant, data-supported recommendation sitting in front of you is almost certainly wrong, because something in the situation doesn't fit and you can't yet say what it is.
That something is judgement.
Judgement is the faculty that operates upstream of every decision you make. It determines whether you're even solving the right problem. It tells you the analysis is correct but the conclusion doesn't follow. It separates the leader who acts on data from the leader who hides behind it.
You use it constantly. You've used it to build a career. But you have almost certainly never been trained in it. No one taught you how it works, where it breaks, what sharpens it, or what quietly erodes it. Especially under pressure. Especially in groups. Especially now, when the tools getting smarter around you can make it dangerously easy to stop exercising it at all.
That's what this masterclass is about.
Not decision-making in the abstract. Not a catalogue of biases. Not tips. This is about the science of judgement: how yours actually works, where it's vulnerable, and how to strengthen it, for yourself and for the people you lead, so that it holds up when it matters. Which is when the pressure is real and the answer isn't obvious.
It takes about 60–90 minutes at your own pace. It moves through three phases. First, we look at tools you already use, ones most people get subtly wrong in ways that cost more than they realise. Then we go deeper, into the traps that no tool can fix on its own, through three organisations that faced them at scale and changed the conditions around their judgement rather than hoping their people would somehow think harder. Finally, we turn to the question most training ignores entirely: how to make sharper judgement stick. Not as an aspiration. As a daily practice.
Let's begin.
Where your judgement stands
Before anything else, find out where you're starting from.
What follows is a diagnostic built on peer-reviewed research. It covers six dimensions of judgement that matter at work. Forty questions, about 10–15 minutes. At the end you'll get a profile: where your judgement is strongest, where it's most exposed, and what that means for the kinds of decisions you face every day.
This isn't a personality quiz. It's the kind of instrument we use with senior professionals and leadership teams, adapted here for individual use. There are no good or bad profiles, only honest ones.
Answer as you actually are. The profile is only useful if it's true.
Once you have your profile, continue below.
The tools you think you know
For most of us, school taught problem-solving, the kind with answers in the back of the book. Judgement, which is reasoning under uncertainty with incomplete information alongside other people, was left to experience. Which means you've been building habits for years, under real pressure, and no one has ever checked whether those habits are good ones.
The fastest progress usually doesn't come from learning something new. It comes from correcting something you already do. Several of the most widely used decision tools are also widely misused, and small adjustments to how you apply them make a disproportionate difference.
Here are two you've almost certainly reached for. The details matter.
The pros and cons list
Joanna is one of those people who makes you recalibrate your definition of a tough job. A uniquely qualified specialist who takes on missions with NGOs in conflict zones, the kinds of places most people only read about. She had just finished a two-year contract in a particularly demanding location and was taking a month to decompress. Her leave was nearly over when she sent me a message: two offers on the table, and she was trying to decide between them.
She was making a pros and cons list.
One of the hazards of being a decision scientist is getting asked for help at the last moment, when someone is tired, under pressure, and facing a choice with real trade-offs and no clean answer. Joanna's situation was the hard kind. She wasn't choosing between two good things. She was choosing between two forms of compromise, and the list she was building was doing what lists do in that situation. Getting longer. More anxious. No clearer.
I asked her to put the list aside.
The pros and cons list is one of the oldest decision tools we have. Benjamin Franklin recommended it in a 1772 letter to the scientist Joseph Priestley, who was agonising over whether to accept a job offer. Franklin called it "Moral or Prudential Algebra." The name shortened over the centuries. So did the method.
Franklin's version was more rigorous than what most of us do today. He advised dividing a sheet into two columns, then spending three to four days, not one sitting, adding considerations as they occurred. Then weighing items against each other: find a Pro and a Con of roughly equal importance and cross out both. Find two Cons that together match one Pro, and cross out all three. Continue until nothing cancels anything else. Whatever remains decides.
His key insight was time. Spreading the process over several days lets considerations surface that aren't immediately salient. The cancellation step forces actual comparison of weight, not just counting. This structured weighing, done slowly and deliberately, was the entire point.
What most of us do today strips nearly everything useful out.
We jot down whatever comes to mind in a single sitting, look at the two columns, and try to get a feeling from the result. The weighing step almost never happens. But the problems go deeper than skipping steps.
The list quietly assumes there are only two options, or at most, two sides. Joanna had two offers, so the list had two columns. But no one had examined whether those were the only possibilities. This is narrow framing: accepting whatever landed on the table first as the entire decision space.
The items on the list are treated as independent when they rarely are. A higher salary and a longer commute aren't separate factors. One may fund the other, or one may erode the other's value over time. A list treats them as separate lines.
Then there's motivated reasoning. If you already have a preference, even one you haven't admitted to yourself, the list will reflect it. You'll generate more cons for the option you quietly don't want and be more generous with the one that appeals. The list feels rational. It's often a tidy presentation of a decision you've already made emotionally.
And perhaps most importantly: a pros and cons list produces all the reasons you think something should be right or wrong, without ever asking whether it's right or wrong for you, and why. It can generate plausible-sounding reasons to pursue choices that don't actually reflect what you need.
Here is where this connects to the world you're operating in right now. Your AI tools can generate a flawless pros and cons analysis in seconds, weighted, scored, cross-referenced. And it will have every one of these same problems. Narrow framing. False independence. No capacity to surface the factors that are hardest to name. The output will look more rigorous. The judgement behind it won't be better. If anything, the polish makes it harder to question.
If you're going to use a pros and cons list, upgrade it.
Widen the frame before you start writing. Are these really the only options? What would a third alternative look like? What would you do if none of these existed? The list works on whatever you feed it. Feed it a narrow frame and you get a narrow answer.
Weigh every item, a number from one to ten, and total each side. Entrepreneur Seymour Schulich adds a useful rule: for a choice to genuinely favour the pros, the positive score should be at least double the negative. That gap corrects for our reliable tendency to overweight the upside. If your pros don't clear that threshold, the decision is closer than it looks.
Add a column for what lists usually leave out. The rational, external factors make it onto the page easily. The internal ones don't: what this choice says about who you're becoming, what it asks you to give up that you haven't named, whether the version of you in five years would recognise the person making this decision. These aren't soft additions. They're data, and often the data that matters most.
Don't finish it in one sitting. Come back over several days, adding, reconsidering, noticing what shifts. Franklin built this into his method for a reason. A list built over time is more honest than one built in a single anxious session.
When I asked Joanna to set the list aside, this is what we did instead. We widened the decision frame. We spent time on the factors she'd left off because they felt too personal to put in a column. And we gave it time. The decision she made wasn't the one the original list pointed to. But she was far more comfortable with it.
Brainstorming in meetings
Shortly after I founded Meta-decisions in 2020, I was invited to meet with a prestigious organisation. The brief was to "brainstorm on ways the team can perform better." For a new founder, that's a meaningful invitation.
I asked for background before the meeting. What projects were they working on? What problems were they trying to solve? The reply was firm: "Just come and we will brainstorm on the spot."
I went. The meeting was fine. But it was nowhere near as useful as it could have been.
This happens constantly, and not only with outside consultants. People within teams face the same dynamic every day. We call a meeting. We expect ideas to emerge. We don't ask whether the conditions for good ideas are actually in place.
Brainstorming became popular in the 1950s with a clear promise: groups produce more and better ideas than individuals working alone. The promise was never quite kept. No study has shown that group brainstorming outperforms individuals working alone first and then coming together. On average, individuals generate more ideas. The meeting, as it's typically run, is not the asset we think it is.
Here's what goes wrong.
In a typical brainstorming session, people overwhelmingly share information everyone already knows and systematically withhold the unique information only they hold. The meeting produces a polished version of what the group already believed. The genuinely new material, sitting in someone's head, never makes it into the conversation.
Whoever speaks first sets the frame. Every idea that follows is evaluated against that anchor, consciously or not. If the first idea is mediocre, the session trends mediocre. If it's the boss's idea, the problem compounds.
The room isn't equal. Extroverts speak more than introverts. Senior people speak more than junior ones. People who are tired, anxious, or simply less verbally dominant stay quiet. Not because they have less to contribute, but because the format doesn't make space for them.
And the people who know a problem most intimately find it hardest to step outside what they know and see it fresh. They've lived with the constraints so long that alternatives obvious to an outsider never occur to them.
That's what I walked into. The team knew their problems intimately. They didn't share the context. Without it, I couldn't see what they were too close to see. We both lost.
Now add AI to this room. An AI-produced briefing document framing the discussion before it starts. An AI tool generating ideas alongside your team. The anchoring effect doesn't shrink. It gets worse. An AI-generated framing carries the authority of "the data says," which makes it even harder for the person with the dissenting insight to speak up. The meeting becomes more efficient at converging on the wrong answer.
The fix starts before the meeting, not in it.
If you're bringing someone in, from inside or outside your organisation, give them the context. What's the actual problem? What are the parameters? What solutions have already been considered? The time spent briefing is returned tenfold.
For the meeting itself, the single most important shift is separating idea generation from idea evaluation. These are different cognitive tasks. Doing them simultaneously, which is what traditional brainstorming demands, compromises both.
The most powerful version of this I've seen in practice: everyone writes for five minutes before anyone speaks. No discussion, no peeking. Then share. That single move removes anchoring, levels the room between introverts and extroverts, and captures unique knowledge before group dynamics can suppress it. Hal Gregersen at the MIT Leadership Center takes it further with the Question Burst method, where instead of generating solutions, you spend four minutes generating nothing but questions about the challenge. No answers, no justifications. The constraint is uncomfortable and productive. Four out of five times, a question surfaces that reframes the problem entirely.
Tony McCaffrey's BrainSwarming replaces talking with writing on a structured graph. The goal at the top of a board, known resources at the bottom, the group working in silence, adding notes and drawing connections. Where the top-down thinking meets the bottom-up thinking is where the solutions live. The numbers are striking: up to 115 ideas in 15 minutes, compared to roughly 100 per hour in traditional brainstorming. Silence turns out to be more productive than conversation.
One last thing about meetings. They tend to suffer not because everyone behaves poorly, but because one person does. One dominant voice, one anchoring comment, one person deferring to the most senior person in the room. Good meeting design isn't optimistic about human nature. It's built to work even when someone has an off day.
What these two tools share
Both widely used. Both subtly broken in ways that compound over time. And in both cases, the fix isn't a new tool. It's a small adjustment to how you use the one you've already got.
This is the first layer of sharpening your judgement: the mechanical corrections. Places where your process has drifted from what actually works, and a specific tweak brings it back.
But correcting familiar tools is only part of the picture. There's a deeper layer, one where the problem isn't the tool you're using but the way your mind works against you before you've even reached for one.