Last week, the Financial Times reported that a KPMG AI-assisted report contained hallucinated claims and bogus case studies.
Not a startup experimenting with AI. Not a small firm cutting corners. KPMG.
And the reason it happened wasn't that the AI was bad. It's that nobody owned the judgment.
We hear that phrase constantly now. "Humans own the judgment." It's become the catch-all reassurance that there's still a human in the loop, that someone is watching.
But here's what nobody is actually asking: what does owning the judgment mean in practice? How does an employee know if they've done it successfully? And when did we last actually develop that skill in anyone?
The Phrase That Sounds Like an Answer But Isn't
"You own the judgment" gets said the way "use your common sense" gets said. As if it's self-explanatory. As if people either have it or they don't.
But judgment is not a personality trait. It's a skill. And like every skill, it has to be defined, taught, and practiced before you can hold someone accountable for it.
If you asked your team right now what "owning the judgment" means in their specific role, what would they say? Because if they can't articulate it, they're not doing it. They're reading the output, nodding, and moving it along.
Which is exactly what happened at KPMG.
What Judgment Actually Means
Owning the judgment means being able to answer yes to all of the following before anything goes to a client:
- Does this make sense? Not just does it read well - does it actually hold up given what I know about this client, this industry, and this situation?
- Can I stand behind every claim? Not "the AI said it" - can I verify it independently and put my name on it?
- Does anything feel off? Even if I can't immediately explain why - have I investigated that feeling rather than overriding it?
- What's the worst case if this is wrong? Have I thought about that before sending it?
That's judgment. It's not a vibe check. It's a deliberate, practiced discipline of critical thinking applied to real output before it reaches a client.
The Training Nobody Is Doing
We have trained people on how to use AI tools. We have not trained them on how to think about the output once it arrives.
In the past, judgment developed slowly through experience. You made a mistake, a senior person caught it, you learned. AI has compressed timelines so dramatically that we no longer have the luxury of waiting for judgment to develop on its own. Output is faster, volume is higher, and the mistakes when they happen are bigger and more visible.
Which means judgment has to be taught deliberately. Not assumed. Taught.
What Judgment Training Actually Looks Like
It starts with defining what good judgment looks like in each role. Not generically - specifically. What does a consultant owning the judgment on an AI-assisted report actually do before it goes to a client? What does an analyst owning the judgment on an AI-generated model actually check?
Then you practice it. You take real AI output and work through it together. What would you check here? What feels off? What assumption is buried in this paragraph? What would you need to verify before you'd put your name on this?
You make the invisible thinking visible. And you do it regularly enough that it becomes instinct.
How You Know Someone Has Done It Successfully
Someone has owned the judgment successfully when they can explain not just what the output says, but why it's right. When they've identified the assumptions and verified them. When they noticed something that felt slightly off and investigated it rather than letting it through.
Not "I checked it." Checked it how? Against what? With what level of scrutiny?
KPMG almost certainly had a process that said humans review AI output. The process wasn't the problem. The standard was.
The Boutique Advantage
Here's what the KPMG story actually proves: the value didn't disappear when AI arrived. It moved from doing the work to verifying it. From generating the output to owning it.
And that is where boutique professional services firms have a genuine advantage over the giants - if they take it seriously.
A large firm with thousands of people and pressure to move fast will always struggle to maintain judgment standards at scale. A boutique with a smaller, more developed team that has been deliberately trained to own their output can deliver faster than ever before and more reliably than firms ten times their size.
But only if they do the work of actually developing judgment as a skill.
Because the firms that supervise the machine will always beat the firms that just trust it.
KPMG just proved that. Expensively.
P.S. - If you're not sure whether your team knows what owning the judgment means in their role, that's your answer. The conversation needs to happen before the mistake does.
FAQs
What does 'owning the judgment' actually mean in practice?
Owning the judgment means being able to answer yes to four things before output reaches a client: Does this make sense given what you know about the client and industry? Can you stand behind every claim independently? Does anything feel off and have you investigated it? Have you thought through the worst case if you're wrong? It's not a vibe check or a surface read. It's a deliberate, practiced discipline of critical thinking applied to real output before it goes anywhere.
Why didn't KPMG catch the hallucinations if they had a human review process?
The process existed but the standard didn't. KPMG almost certainly had a requirement that humans review AI output. But nobody had defined what that review should actually look like or what level of scrutiny was required. So people read the output, saw it looked reasonable, and moved it along. A vague process with no clear standard fails at scale, especially when output volume and speed are high.
How do I know if my team actually has judgment about AI output?
Ask them what 'owning the judgment' means in their specific role. If they can't articulate it clearly and specifically, they don't have it yet. You need them to be able to explain not just that they reviewed something, but how they reviewed it, what they were checking for, and why they'd put their name on it. The inability to articulate this is the problem that causes expensive failures.
Can judgment be trained, or is it something people are born with?
Judgment is a skill, not a personality trait. Like any skill, it requires definition, teaching, and deliberate practice. The old way was to learn through experience over time by making mistakes and having them corrected. AI has compressed timelines so dramatically that we can't wait for judgment to develop slowly anymore. It has to be taught deliberately by making the thinking process visible and practicing it on real output regularly.
What's the competitive advantage for smaller firms in this situation?
Boutique professional services firms can deliver faster than giants while maintaining reliability that scaled organizations struggle to match, but only if they deliberately develop judgment as a skill. Large firms with thousands of people under pressure to move fast will always struggle to maintain consistent standards at scale. A smaller, more developed team that's been trained to own their output can catch what the machine might miss before it becomes a crisis. The firms that supervise the machine beat the firms that just trust it.



