Developing leaders in the age of AI means building judgment on purpose, because the work that used to build it by accident is being automated. That was the question underneath the executive panel I joined at the inaugural Chief Seattle Salon this September. It's also the one I keep hearing from the CHROs I talk to.
Every woman on that stage built her judgment the slow way. We spent years inside the manual, unglamorous version of the work. We sat with the numbers or the client or the problem long enough to recognize when an answer was wrong. So, here's the question I'd put to every people leader right now: if that version of the work is no longer on offer, where does the next generation's judgment come from?
The correction loop is disappearing. AI is absorbing the low-stakes work where people used to get things wrong, get corrected, and learn.
Discernment isn't judgment. AI is very good at knowing the rule. Only a person can know when this situation is the exception.
Order matters. Do the human work before you buy the tool, then design judgment-building into the job instead of hoping it happens.
What used to build judgment was getting something wrong in front of your boss, at low stakes, a hundred times before anyone trusted you with a real call. That loop is going away. In a 2025 McKinsey survey, 51 percent of organizations said generative AI was reducing their need for entry-level roles. The team that used to be five analysts is now one person and a tool.
I'm seeing the effect in coaching. Leaders are stepping into bigger roles faster than the traditional reps would have justified, and it isn't because they lack talent. One leader I work with moved into a Chief Transformation Officer title sooner than that path used to allow, because AI flattened the layer where people used to cut their teeth. The promotion happened on paper before the judgment was built underneath it.
I'm seeing the effect in coaching. Leaders are stepping into bigger roles faster than the traditional reps would have justified, and it isn't because they lack talent. One leader I work with moved into a Chief Transformation Officer title sooner than that path used to allow, because AI flattened the layer where people used to cut their teeth. The promotion happened on paper before the judgment was built underneath it.
Here's where it shows up. People are looking at clean, confident AI output, and they don't have the scar tissue yet to know when confident is wrong. I've sat across from leaders who can execute beautifully and can't tell me why they made the call, because they didn't really make it. The tool did, and nobody ever made them defend it.
That's leadership debt, right? You defer the investment in how people learn to decide, and you keep operating. Then the bill arrives in somebody's first ninety days in a job they weren't built for yet. McKinsey makes the same point from the pipeline side: skipping entry-level hiring risks weakening the future leadership pipeline.
Discernment is knowing the rule, and AI is very good at it. Judgment is knowing that this person, in this moment, is the exception to the rule.
That distinction matters because judgment runs on information that's never written down anywhere. It lives in a raised eyebrow in a meeting, a slower reply to an email than usual, the thing someone doesn't say out loud. That's why I treat empathy as a data-collection tool rather than a soft skill you add after the hard decision gets made.
Say I'm sitting across from someone whose role I'm about to change significantly. I'm watching whether they're meeting my eyes and whether they seem confused or upset. That tells me whether to slow down, and whether the version of the conversation I planned in the car on the way to work is the wrong one for this person. AI produces a script and assumes you followed it. It can't recalibrate mid-sentence because of what's happening in the room.
|
|
Discernment |
Judgment |
|
What it is |
Knowing the rule |
Knowing when this case is the exception |
|
What it runs on |
Documented data, patterns, and precedent |
Context no one wrote down: tone, history, timing |
|
Where AI stands |
Strong, and getting stronger |
Can inform it, can't supply it |
|
How it gets built |
Training and exposure |
Reps, correction, and having to defend the call |
Three categories stay human by design in the work I do:
performance conversations
anything with a title or reporting line attached
any decision where you have to read a room instead of a spreadsheet
AI can hand you the org chart that optimizes for span of control. It can't tell you that someone is one bad Monday from quitting. It can't tell you a team is underperforming because two people on it haven't spoken up since a reorg two years ago. Those are exactly the facts that can change the right answer.
We used AI this year to model scenarios for an org redesign. We didn't let it draft what people were told, and we didn't let it decide who delivered the news to whom. Those two calls, who says it and how, are entirely about reading one specific human being.
The same goes for individual decisions. Recently, a scorecard told me to cut a long-tenured underperformer. What the scorecard didn't have was that he'd trained half the people still on the team. This needed to be a conversation about his whole career, not a line item, so I made that call myself.
Because order matters, and a lot of companies get it backwards. Change management happens before you buy the tool, not after. Gartner found that 78 percent of CHROs agree workflows and roles will need to change to get the most out of their AI investments. That redesign doesn't happen on its own once the software goes live.
Two things make it real. First, a readiness check before anyone spends a dollar: does this team even have a documented process to hand to a system, or does the knowledge live in one person's head? Second, work redesign, instead of a new tool bolted onto the old job description.
Skip those and you get digitization, not transformation. People can feel the difference immediately.
|
|
Digitization |
Transformation |
|
What changes |
A new tool is added to the existing workflow |
The workflow and the role are redesigned |
|
What happens to the old work |
It stays, with one more step on top |
Low-value work falls away |
|
How people experience it |
More to do, same expectations |
Time back for harder, human calls |
|
Where judgment lands |
Deferred to the tool by default |
Assigned to people on purpose |
The pushback I hear most is “we don't have six weeks for readiness work, we need this live.” That's usually the company that calls back four months later asking why nobody's using it. Skipping the step also costs more than it looks. Gartner reports that only 1 percent of layoffs in the first half of 2025 came from AI actually making employees more productive. Organizations are cutting on returns that haven't arrived yet.
That sequence is what livingHR's AI-People Solutions work is built on: a readiness assessment first, then org design, workflow optimization, and learning and change, in that order. For a broader view of where the function is headed, see How AI Will Transform Your HR Function in 2026.
Stop treating judgment-building as a side effect of doing the job long enough. It used to be one. Now it has to be designed.
One small lever I use: I stopped letting coaching clients bring me a conclusion. If someone shows up with an AI-drafted recommendation, they argue the other side first, out loud, before I'll engage with the recommendation itself. If they can't, we're not building judgment. We're rubber-stamping a draft.
That practice travels well beyond coaching. A few questions worth asking inside your own organization:
Where does someone early in their career still get to be wrong at low stakes, and get corrected?
When AI drafts the recommendation, who has to defend it, and to whom?
Which decisions have you assigned to a person on purpose, and does everyone know which ones those are?
My stop and start from the panel still holds. Stop buying the tool before you've done the human work. Start designing for judgment the way you'd design for any other capability you can't afford to lose.
Nobody in that Seattle room had the human and machine partnership fully worked out, and I'd be a little suspicious of anyone who says they do. So I'll hand the question back to you. What's one decision in front of you right now that you'd approach differently if you designed for judgment first? It could be a tool, a restructure, or a role you're about to define.
If you're sorting out which decisions to keep human as AI moves deeper into your organization, that's work we do with people leaders every day at livingHR. Let's talk.
AI is automating much of the entry-level work where people used to build judgment through low-stakes mistakes and correction. Leadership development now has to recreate that correction loop on purpose.
Discernment is knowing the rule, which AI does well. Judgment is recognizing when a specific person or situation is the exception, and it depends on context that's rarely written down.
Performance conversations, decisions involving titles or reporting lines, and any call that depends on reading people rather than data. AI can inform these decisions, but a person should own them.
Before. A readiness check and work redesign up front are what separate real transformation from adding one more tool to an old workflow.
Make them defend the call. Asking someone to argue the other side of an AI-drafted recommendation before acting on it turns a rubber stamp back into a decision.