PwC asked 1,004 financial services executives what they want from the people they hire. 86% agreed that AI skills training is more valuable than an MBA for many new hires.
That figure ran everywhere as a threat story. I think it is closer to the opposite, and the reason sits in what kind of thing an AI skill actually is.
An MBA takes two years and costs somewhere north of £60,000. It is a credential, and credentials work on a schedule you cannot rejoin. If you did not get one at the right point in your career, that door is mostly shut, and no amount of being good at your job reopens it.
AI fluency is not that. It is a practice. It compounds with the expertise you already have, and it can be acquired in months by someone who has been working for thirty years. The survey is telling you that the market has started paying for the second kind of thing rather than the first.
For anyone who has quietly concluded they missed the boat, that is the most useful piece of news this year.
What the survey actually found
Three figures matter, and they should be read together.
86% agree AI skills training is more valuable than an MBA for many new hires. Keep the “for many new hires” attached. This is a statement about what employers screen for at the front door, not a claim that MBAs have stopped meaning anything.
91% say they are increasing compensation for employees with AI skills. This one is not about new hires. It is about the people already there.
58% say they will tie compensation directly to AI-enabled productivity. Note the wording, which is PwC’s own. Some coverage rendered this as firms being “willing to pay a premium for AI fluency”, which is a softer and different claim. Tying pay to AI-enabled productivity is a structural change to how these firms intend to reward work.
The bridge between the first figure and the other two is the part worth holding onto. What gets screened for in new hires is a signal. What gets paid for across the existing workforce is the actual market, and it is moving in the same direction.
Where this applies, and where it does not
PwC surveyed 1,004 executives at director level or above, at US financial services firms with at least $500 million in revenue, between 12 and 22 May 2026. The sample splits evenly across asset and wealth management, banking and capital markets, insurance, and private equity.
That is a narrow, large, American slice of one sector. If you work in a forty-person firm in Manchester, this is not a description of your market. It is a leading indicator from the part of the economy that usually moves first on this kind of thing, and typically two or three years ahead of everyone else.
Read it that way and it is still useful. Read it as a description of where you already are and it will make you panic unnecessarily.
The part I am not going to skip
The same survey found that nearly eight in ten of those executives expect their workforce to shrink by at least 20% over the next five years. Entry-level roles were named as most exposed by 30% of them, with middle management next at 26%.
That is a genuinely difficult finding and it deserves to be stated plainly rather than buried.
Two things sit alongside it, though, and both are from the same survey. 77% said most of their AI investments are not currently delivering measurable return. And 34% named change fatigue as a key barrier to scaling AI across their workforce.
So this is not a smooth march toward a smaller workforce. It is a set of expectations held by people who also report that the thing is not paying off yet and that their organisations are tired. Expectations at five years are the least reliable numbers in any survey.
What “AI skills” actually means here
Not programming. Almost none of what these firms are paying for involves writing code.
What they are describing is fluency: knowing what to hand to a model and what to keep, being able to tell a good output from a plausible one, recognising when the answer is confidently wrong, and building the judgement to know which parts of your own work benefit and which do not.
That is a professional skill, and it sits on top of domain expertise rather than replacing it. PwC’s own AI Jobs Barometer points the same way: roles that AI makes more expert rather than more accessible are growing twice as fast by volume and showing 42% higher wage growth.
Which is to say the thing being rewarded is not “can operate ChatGPT”. It is “knows this field and can also do this”. Thirty years of knowing a field is not a disadvantage there. It is most of the asset.
What to do about it
Treat it as a practice problem rather than a credential problem, because that is what it is.
Nobody is going to award you an AI qualification that changes how you are paid. What changes is whether you can sit down with a real piece of your own work, use a model on it properly, and tell whether the result is any good. That takes protected time and repetition, and it takes doing it on work that matters rather than on demos.
If you have been putting off getting fluent because it felt like a young person’s game, the survey is quite clear that it is not. It is an experienced person’s game with a new tool in it.
This is the week to start.
Source: The AI workforce planning gap in financial services, PwC, 3 August 2026. Survey of 1,004 executives, 12 to 22 May 2026.