On 5 February 2026 PwC US launched the Learning Collective, a firm-wide model for developing what it calls an AI-native workforce. It rests on 30 named skills: 15 AI technical skills and 15 human skills, among them critical thinking, judgement, analytical reasoning, intellectual curiosity, adaptability, empathy and the ability to explain a conclusion to someone more senior. It comes with a rule, set out by Yolanda Seals-Coffield, PwC US Chief People and Inclusion Officer: anywhere in the firm that an AI skill is taught, a human skill is taught with it, and the two are not to be taught apart. The model follows pilots run through 2025, moves learning out of classrooms and courses and into client work, and will be rolled out across the United States this year. PwC says its learning programmes will be evaluated for effectiveness and relevance, not just completion. Six weeks later, on 19 March, PwC US senior partner Paul Griggs told the Financial Times that anyone who thought they had the opportunity to opt out of AI would not be at the firm for long. The PwC network employs more than 364,000 people in 136 countries. Alongside the Learning Collective, its Human Skills Project has committed one million service hours to reach 250,000 learners outside the firm over three years.
That is the announcement. Here is why it matters more to a professional body than to PwC’s competitors.
Read the list again
Fifteen human skills. Judgement. Critical thinking. Scepticism about an output. The ability to reason from evidence and defend the reasoning to a partner. Curiosity about why a number looks the way it does.
A professional body will recognise every line, because every line has been in its own competence framework for as long as it has had one. The chartered designation has always rested on judgement rather than on knowing how to work a tool. Tools change; judgement is what the letters after a name are supposed to certify.
What is new is not the content. What is new is that an employer has written those qualities down against AI specifically, given them a number and a date, put them on the same page as prompting and responsible AI, and is teaching them every working day, inside the work, to the people it will decide to promote or not.
Last week I wrote that the employer teaches the tool and the profession owns the duty. PwC has not tried to take the duty. It has done something a body should look at more carefully. It has written its own working definition of what good judgement with AI looks like, and for its people that definition is now the one that counts.
What PwC’s list means for a professional body’s competence framework
A body’s competence framework is written to last. That is what a standard is for. It goes through committees and consultation because it has to hold for tens of thousands of members across every kind of employer, and because a member qualified under it in 2019 must still be able to point to it in 2029.
PwC’s list was written in about a year, tested in pilots, launched, and will be rewritten when the tools move. That is what a firm’s list is for.
Neither document is wrong. The risk is what happens between them.
When a member says they know how to use AI properly, the follow-up question is: by whose definition.
For a trainee in a large firm today the honest answer is the firm’s. The fast, local document becomes the working definition because it is the one they are measured against on Monday. The slow, universal one stays on the shelf.
Nobody at PwC set out to displace anything. They solved their own problem well. That is exactly why it will be copied, and why the working definition of professional judgement with AI is being written, this year, by employers.
Three things a firm’s list cannot do
It stops at the door. The Learning Collective covers PwC’s people while they are at PwC. Most trainees leave, and when they go, the firm’s development record stays behind. What travels with them is the designation. The next employer, who may be a forty-person practice with no learning function at all, reads the letters after the name and assumes they mean something about judgement in the current world of work. Whether they do is the body’s decision, not the firm’s.
It is not independent. A firm cannot certify to the public that its own people’s judgement meets a standard the firm itself wrote. It can tell a client what it trains for. It cannot be the assurance. That independence is the whole reason a chartered designation exists, and it is not something PwC or anyone else can manufacture internally.
It covers a fraction of the profession. Most members of any body work in organisations that will never write a 30-skill framework, run a pilot programme, or pair an AI skill with a human one on purpose. For those members, the only place that judgement with AI will ever be defined, taught or evidenced is the body they belong to. If it is not defined there, it is not defined anywhere.
The employer’s move creates a gap, and the body is the only institution shaped to fill it.
How a professional body can teach AI and judgement together
Set the 30 skills aside for a moment. The most transferable thing PwC has published is the rule: an AI skill is never taught without a human skill in the same session.
Look at what that rule forbids. It forbids the shape most professional CPD programmes take when AI arrives: a course on the tool in one place, an ethics module in another, a completion recorded for each, and the member left to join them up alone, at their desk, with a client waiting. The two things a member most needs to do together are taught apart.
To teach judgement alongside an AI skill, you have to put the member in a situation where an AI has already produced something, and they have to decide what to do with it. Check it, disclose it, hold it back, or send it. That is one experience, not two. The tool is the setting; the judgement is what gets examined.
That is a design rule, not a purchasing rule, and it can be applied to a professional programme of any size. It costs nothing to adopt and it changes what the member walks away having practised.
Two moves a professional body can make this quarter
Two practical moves follow, and neither requires a committee cycle.
First, take the competence framework off the shelf and read it with one question: where does it say, in words a member could be assessed against, what good judgement with AI looks like? If the answer is nowhere yet, the employer’s document is the reference document by default, and the body’s members are being measured against a standard the body did not write.
Second, look at the AI content the body already offers or plans to commission, and apply PwC’s rule to it. Where the tool and the judgement are taught in separate places, put them in the same room.
LearnFrame built AI for Professional Practice on exactly that rule. Every decision in the scenario is a decision about something an AI has already produced for the member. Nothing is scored on whether they can operate the tool; everything is scored on what they choose to do with its output. Module 1 is live, it takes twenty minutes, and it runs on a phone. Any education team can complete it today at learnframe.com/course/ai-for-professional-practice/module-1/, and it can be licensed under a body’s own name, so that the working definition of professional judgement with AI carries the profession’s mark rather than an employer’s.
PwC has shown what the employer’s half of this looks like when it is done properly. The half that travels with the member, and holds across every employer, was always the body’s to write.
Want to know what your own programme could tell you that it currently does not? The Programme Design Diagnostic is a fixed-fee, independent read on a single programme: a board-ready diagnosis and a costed build scope, in two to three weeks.
Sources: PwC US, “PwC Learning Collective Reimagines Development for the AI Age”, press release, 5 February 2026, including the 30 skills, the 2025 pilots, the United States rollout, the 364,000 people in 136 countries figure and the Human Skills Project commitment. Charter, “How PwC is changing learning for the AI era”, interview with Yolanda Seals-Coffield, 5 February 2026, for the rule that AI and human skills are taught together and the named human skills. HR Grapevine, 5 February 2026, on evaluation for effectiveness and real-world relevance rather than completion. Financial Times interview with Paul Griggs, 19 March 2026, as reported by Accounting Today, CPA Practice Advisor and The Register. All retrieved 16 September 2026.
For the two layers of AI training and who owns each, see Santander (2026): AI tools for all 185,000 staff, mandatory AI training, and a €1 billion target. For another large employer’s AI training at scale, see Accenture’s AI upskilling programme (2026): 550,000 staff trained. For what AI literacy means in plain terms, see AI literacy in plain English. For the module itself, see AI for Professional Practice: Module 1 is live. See more insights from LearnFrame.