AI Is Making Organisations Mediocre
Not because the models are dumb. Because they made sounding reasonable almost free
No matter what role you are, the lifebuoy you reach for is the same one: being seemingly reasonable. It keeps you afloat. It also guarantees you never swim anywhere.
I was at a company offsite a few weeks ago, and somewhere between sessions I got talking to my boss about how he actually spends a day. The thing that stuck with me: he makes dozens of decisions a day. Not big strategic ones each time. Mostly a steady stream of this, not that, over and over. For him it’s reflex now. He doesn’t agonise. He ranks, picks, moves on, and lives with it.
I keep coming back to that conversation, because it points at the one skill AI is quietly making rare. Not knowledge. Not analysis. Judgment - and specifically the nerve to choose.
For someone like my boss, judgment is instinct. He built it over years of having to decide with half the information and own the result. The problem is that most people haven’t built that muscle yet, and the thing that used to force them to is disappearing.
So this matters most for the people who never had it drilled into them as instinct. The environment used to do the drilling. It’s stopping.
The thing that got cheap
Picture a meeting from a few years ago. People show up, a few of them offer a real angle; we are lucky if we leave the meeting with 3 really good ideas; and that counts as a serious discussion. Perspectives were scarce. You worked to produce one.
Now everyone walks in armed. AI organises your point, fills in the risks you’d have missed, drafts the plan, tightens the wording. Nobody is short of an opinion anymore. The meeting is full.
What’s missing is judgment.
1. A surplus of correct, useless opinions
Marketing says we can’t sacrifice user experience. Sales says we have to respond to what customers are asking for. Engineering says complexity can’t keep climbing. Finance says cost has to come down. Legal says risk has to be contained.
Every one of those is true.
And now each of those people can use AI to make their slice of the truth more complete, more polished, harder to argue with. The org ends up with more viewpoints than it has ever had.
More viewpoints isn’t more wisdom. It’s just more raw material. Judgment is the thing that turns material into a structure. Without it you’ve got a very expensive pile.
2. Hearing everyone out flattens everything
A technical problem, a process problem, an org problem, and a values problem are not the same size. They don’t sit on the same level.
But the moment they hit the meeting notes, they line up as bullet points of equal weight. This needs optimising. Watch out for that. Consider this risk too. That point was important as well.
Everyone got heard. Nothing got seen clearly.
Real judgment is knowing which of those points is closest to the thing actually driving the problem - and saying so out loud, while the rest are still nodding.
3. The opinion that matters is rarely the one most people repeat
There’s a trick the room plays on you. An opinion that several people echo starts to feel like consensus. An opinion only one person raises feels like a footnote.
But the judgments that actually move things are usually the ones only a couple of people can see. They don’t add a tidy piece to the existing frame. They change what the frame is about.
That kind of point tends to land badly. It’s not loud. It’s awkward to pick up and run with. It can sound slightly off-topic. And it’s often the one touching the variable underneath everything else.
With no judgment in the room, that point gets buried under a stack of “yeah, that’s also fair.”
4. AI is great at exactly the wrong thing
Ask what AI is best at. Synthesising. Balancing. Filling gaps. Smoothing a document until it looks finished.
Which means it’s a perfect crutch for an organisation that doesn’t want to commit:
A team that’s scared to choose will use AI to generate more options. A team that’s scared to call it will use AI to lay out more pros and cons. A team that’s scared to upset anyone will use AI to write something more carefully worded. A team with no through-line will use AI to produce a prettier composite of everyone’s input.
What comes out the other end looks mature. Context, goals, a path, the risks, who owns what, the milestones. It has everything.
Except a single decision that someone is on the hook for.
5. Mediocrity is getting better dressed
Mediocre work used to look mediocre. Rough, muddled, half-thought-through. You could spot it.
The AI-era version is harder to catch. It’s complete, sensible, professional, structured. It reads like a finished plan.
Then you get to the end and notice what isn’t there. No direction anyone is actually betting on. Not one line that carries real risk. Nobody saying: park these nine things, we’re only doing this one.
That’s the new shape of it. High polish, zero judgment. Mediocrity that passes review.
6. Judgment shows up first as the nerve to rank things
Judgment isn’t the loudest voice in the room, and it isn’t the most senior title. It starts as the ability to put things in order. To look at a pile of perfectly reasonable opinions and know which is the spine and which is the noise. Which is the variable underneath and which is just locally true. What has to be dealt with now and what can wait. What’s worth upsetting people over and what cost is worth carrying.
This is never comfortable. Ranking means leaving things out. Leaving things out means some reasonable point gets set down. And someone walks away feeling their input didn’t count.
So plenty of organisations quietly decide they’d rather not. They pick discussion, alignment, balance. They end up with a version nobody can find a real fault with — which looks like maturity and is mostly just a more dignified way of not deciding.
What actually separates teams now
AI doesn’t make an organisation mediocre on its own. It takes the teams that were already bad at deciding and lets them produce mediocrity faster and prettier.
The gap that’s going to open up between teams isn’t about who uses AI. Everyone will. It’s about whether, after the model hands you ten plausible directions, anyone in the room will stand up and pick one.
AI can generate the many. Only judgment turns the many into order.
And this is where my boss’s point lands for me. He makes those calls all day without thinking, because the old way of working forced him to build the instinct early - back when options were scarce and you had to choose. The people coming up now don’t get that pressure. The tools hand them completeness before they’ve ever had to commit to anything.
If you’ve already got the instinct, AI is leverage. If you haven’t, AI is a very comfortable place to hide.
That’s the uncomfortable part. The one skill that just got scarce is the exact thing you can’t hand to the model, and the better the tools get, the easier it is to put off building it. Comfort is the trap. You can ship flawless, thorough, well-reasoned work for years and never once decide anything.
So act on it. Not someday - in the next meeting where ten reasonable options are on the table, be the one who says which matters and which can wait. Pick, out loud, with your name on it. Be willing to be wrong where people can see. (That part never gets easier. You just stop minding it.) That’s the rep, and it’s the only way the muscle gets built, because nothing in your day is going to force it on you.
A meeting full of well-argued opinions with nobody willing to rank them doesn’t get you somewhere smarter. It gets you somewhere very neat, very thorough, and going nowhere in particular.
Judgment is the moment someone lets go of the ring and picks a shore to swim for. Colder. Riskier. The only thing that ever gets you out of the water.