What a Certificate in Judgment Can and Can't Teach
This month PwC US announced that it will put more than 85,000 partners and employees through a new Critical Thinking & Judgment Certificate. According to Bloomberg Tax and Business Insider, the program opens with a diagnostic of each person's thinking style. From there it moves to masterclasses and personalized simulations, and it adds an AI coach trained on PwC's own framework. The firm estimates the total investment at more than $100 million. That figure includes funding for selected universities to build introductory courses around the framework. Completion is targeted for mid-2027, and coverage says it will be tied to performance reviews.
It would be easy to roll your eyes at this. A big firm buys a big certificate and announces it with a big number. But the reason PwC gives for the program is honest, and it deserves a hearing before any criticism.
The problem PwC is naming
PwC's leadership has said, in effect, that AI is taking over the routine work that used to teach people judgment. That is a real problem, and it isn't unique to accounting firms.
Think about how anyone gets good at spotting a bad number. You build hundreds of them yourself. You get some wrong, someone more experienced catches the mistake, and you remember how it felt. After enough rounds, something looks off to you before you can say why. That instinct came from repetition and correction, not from a class.
When AI does the building, the repetition disappears. The output still needs someone to check it, because AI tools still make confident mistakes. They misread a source, invent a figure or apply a rule that doesn't fit, and they sound just as sure when they're wrong as when they're right. So the work now calls for more skill at reviewing, while the job no longer gives people the practice that builds that skill.
Why "AI will fix it" doesn't close the gap
The common answer is to let AI check AI. Sometimes that works. When there's an outside test, like code that either runs or doesn't or totals that either tie out or don't, a second pass can catch real errors, because the test decides what's right.
The hard questions in business usually don't come with a test. Is this assumption reasonable? Does this contract term actually protect us? Does this forecast pass the smell test? For questions like these, a second AI pass means asking the same kind of system to grade its own thinking. It tends to repeat the same blind spots, with the same confidence. A review like that feels thorough without being one.
PwC seems to understand this, and that's why the program exists. The question is whether a certificate is the right tool.
What a certificate can do
A well-built program can do several useful things. It can give a large organization a shared vocabulary, so a manager can say "walk me through your evidence" and everyone knows what that means. It can turn good habits into a routine, such as checking the source, asking what would have to be true, and looking for the number that doesn't fit. Simulations add practice reps in a setting where mistakes are cheap. A firm-wide program also sends a clear message that leadership expects people to question the tools rather than simply trust them.
These are real benefits, and for an organization of 85,000 people, consistency alone may be worth a lot.
What a certificate can't do
A certificate can't prove that someone has judgment. Judgment shows up on real work, with real consequences, when no one has written the answer key in advance. A simulation is graded against an answer someone already decided was correct. That's useful practice, but it's a different thing from the work itself.
There's also no public evidence yet that this program works. According to International Accounting Bulletin, PwC developed the framework internally and is validating it with academic input, including professors affiliated with Wharton and Harvard, as it goes. That's a reasonable way to start, but it means the program is a well-funded experiment, not an established standard. Tying completion to performance reviews adds another risk. When a course becomes a box to check, people learn how to check the box.
The university funding deserves a clear-eyed look too. Paying for introductory courses built around one firm's framework helps students, and it also builds that firm's brand with future hires. That isn't sinister, but readers should see both sides of it.
What's coming next
When a firm this size puts nine figures behind "judgment in the age of AI," others will follow. Expect weekend courses, digital badges and new training companies promising to make your team AI-ready or certified in AI oversight. Many will copy the language of a program like PwC's without the scale, the simulations or the senior people behind it.
If one of those lands in front of you, a few questions can help you sort it out. Does it use your own work, or generic cases? Who reviews the output: an experienced person, or another tool? What does it say the certificate proves, and is there any evidence beyond completion rates? Does it assume you still have experienced people to learn from, or does it promise to replace them?
The real test
The skill PwC is trying to protect is worth protecting, and it was never built by a course. It was built by doing the work while someone who knew better looked over your shoulder. That gives anyone a simple way to judge this program, or any of the copies that follow it: ignore the certificate and watch the review. If experienced people spend more time checking AI-assisted work, the program is building judgment. If review time shrinks because everyone is now certified, the firm has traded the people who had judgment for a document saying everyone does.