Leadership

PwC will certify 85,000 US staff in critical thinking to check AI’s work

Written by
Full Name
November 13, 2026
PwC US is spending more than $100m, by Bloomberg Tax’s estimate, on a judgement certificate that from December grades every partner and employee on how well they challenge AI-generated work, with results feeding into performance reviews.

PwC’s response to AI taking on more of its client work is to spend more than $100m (roughly £76m) training its people to argue with it. The firm’s US arm said on 9 November that all of its more than 85,000 partners and employees will complete a Critical Thinking & Judgment Certificate, starting in December and finishing by the end of its fiscal year in June 2027.

Critical thinking is the skill clients assumed they were already paying for. What is new is the formality. PwC is turning it into a firm-wide credential, assessed partly through simulations in which staff pick apart convincing AI-generated work, and feeding the results into performance reviews. For marketing teams that now run more of their own output through AI tools, it is a large, public example of an organisation putting a budget and a measure on the review step instead of treating it as free.

What does PwC’s critical thinking certificate involve?

PwC’s certificate rests on a proprietary Critical Thinking & Judgment framework the firm built itself. It is not an external accreditation, and neither PwC nor Bloomberg Tax names a delivery partner. PwC says it is validating the framework with input from professors affiliated with the Wharton School and Harvard University; the announcement does not say either institution endorses the certificate.

The programme opens with a diagnostic scenario that shows each participant their thinking preferences, strengths and gaps. Staff then learn the framework (establishing context, evaluating information, challenging assumptions and communicating a recommendation), watch masterclass-style videos with PwC leaders and outside speakers, and work through simulations tailored to their roles. Toolkits and team exercises follow to carry the habits into daily work, and HRD America reports an AI coach trained on the framework will help staff practise as they go.

The simulations show what PwC is worried about. HRD America describes one in which a tax professional reviews an AI-generated analysis of a complex position that looks convincing but may rest on incomplete support. The participant has to question the sources, weigh other readings and decide what else is needed before reaching a view. Bloomberg Tax reports a possible exercise built around an AI-generated assessment of a company’s internal controls.

Margaret Burke, PwC US’s talent acquisition and development leader, told HRD America it is “still up to us to question those outputs”. Progress will be judged through a final assessment of how participants apply the framework, alongside observation and performance feedback. PwC has not published what a pass requires. Business Insider reports the certificate is designed to take around six months.

Why is PwC paying to teach a skill it already hired for?

PwC’s own explanation centres on how junior staff learn. The firm says first-year associates already do different work from their counterparts two years ago, because AI now handles the foundational tasks through which people used to build judgement, according to HRD America’s account of the announcement. If the groundwork no longer teaches the skill, the firm has to teach it some other way.

Yolanda Seals-Coffield, PwC US’s chief people and inclusion officer and the author of the announcement, wrote that sound human judgement “will remain essential to earning trust and delivering quality work”. For an audit and advisory business, that is a commercial point as much as a cultural one. Clients pay PwC to stand behind conclusions, and an AI-drafted conclusion that nobody has properly tested is a liability with the firm’s name on it.

The $100m figure needs reading with care. It comes from Bloomberg Tax, which reported it as PwC’s estimate of total investment, including programme development and funding for universities. PwC’s own announcement gives no figure. It is therefore not a per-head training budget, and dividing it across 85,000 people would overstate what each employee receives.

The university element is substantial in its own right. From early 2027, PwC plans to offer the framework to any university that wants to teach it and, through its Student Human Skills Initiative, to fund selected institutions to build introductory courses around it. That initiative extends PwC’s Human Skills Project, to which the PwC Foundation committed $2m (roughly £1.5m) earlier this year.

How does the certificate fit the Big Four’s wider AI training push?

PwC is running the certificate alongside its tool training. In May 2026 PwC and Anthropic expanded their alliance, with PwC committing to train and certify 30,000 US professionals on Claude and to set up a joint centre of excellence. IT Brief reported that Deloitte and KPMG have each committed to certify 15,000 professionals on Claude, and Capgemini 20,000.

Those programmes certify how to use a tool. The new certificate covers the person checking what the tool produces, and it reaches nearly three times as many PwC staff as the Claude programme. The Helm found no comparable firm-wide judgement credential announced by another Big Four firm as of 13 November, though all four are investing heavily in AI adoption.

That has consequences beyond accounting. PwC is one of the larger buyers and implementers of enterprise AI, and it is now training its entire US workforce to treat AI output as a draft to test before it goes anywhere. Vendors selling AI into professional services are pitching to buyers whose staff will be graded on exactly that habit. Product marketing that shows how an output can be traced to its sources, challenged and escalated speaks to what those buyers are being trained to do; claims built only on how much work disappears speak less well to it.

For in-house marketing teams, the review work PwC is formalising is already familiar. An editor who checks an AI-drafted case study against the customer’s own numbers, or a demand-generation manager who asks why a model’s lead score jumped, is doing the same thing PwC’s simulations rehearse. The difference is that PwC has now costed it, built a framework for it and attached it to performance reviews.

The first cohort starts in December. PwC has not yet said what happens to anyone who does not pass.

Subscribe to our newsletter

By subscribing you agree to with our Privacy Policy
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Share article

Recommended Reading