PwC Is Spending $100 Million to Teach 85,000 Staff Not to Trust AI

PwC is putting $100 million behind a certification program that teaches more than 85,000 employees and partners how to doubt AI — after training them to use it. As Bloomberg Tax reported, the initiative centers on critical thinking and judgment in a workplace increasingly run by automated systems.

The program's core skill is pushback. Margaret Burke, PwC U.S.'s leader of talent acquisition and development, gave Human Resources Director a concrete example: an AI model produces a tax analysis. The training teaches staff to trace the model's logical process and identify where it pulled its data from, rather than accepting the output as correct. The employee then decides whether they trust the conclusion or whether it needs more fine-tuning.

"It's an opportunity to practice the judgment involved in making a decision, rather than simply learning about critical thinking in theory," Burke said.

Yolanda Seals-Coffield, PwC's chief people and inclusion officer, framed the program in equally practical terms in a press release: "Our goal is to give our people a practical, repeatable approach to evaluating information, challenging assumptions, and exercising sound judgment—particularly when AI is involved."

Why is a consultancy paying people to distrust its tools?

Because the evidence says AI use erodes exactly the skill PwC sells. One recent study found that higher AI usage correlated with lower critical thinking due to "cognitive offloading" — the phenomenon where offloading a skill to a machine makes you worse at that skill yourself.

The pattern has historical precedent. GPS usage degraded people's sense of direction. Smartphones weakened the ability to remember phone numbers. Scale is the difference this time: generative AI can absorb far more of our thinking than either of those technologies ever did.

"One of the things that's seductive but also troubling about generative AI is the ability to offload much more executive functioning than we have in the past," said Nita Farahany, a philosophy professor, in an interview with the American Psychological Association.

Farahany posed the stakes plainly: "What does it mean for society if humans are passively receiving information but no longer able to critically interrogate it?"

What does this change for practitioners?

For a firm whose entire product is judgment — tax analysis, audit conclusions, advisory recommendations — the program is a defensive investment. If clients can't distinguish a PwC analysis from a chatbot's, the premium collapses. Training 85,000 people to interrogate AI outputs keeps the human layer accountable for the final call.

The $100 million figure, spread across a workforce of that size, works out to roughly $1,175 per head — a signal of how seriously the firm rates the cognitive-offloading risk.

The move also sets a template other large employers may follow. If AI-generated first drafts become standard, the scarce skill becomes verification: knowing when the machine's reasoning holds, when its sources are shaky, and when human judgment has to take over entirely.

PwC is betting that skepticism, not fluency, becomes the defining professional competency of the AI era.