What Happens to Your Firm's Expertise When Your Best People Leave

PwC's 2026 Global AI Jobs Barometer, an analysis of more than a billion job advertisements across 27 countries, found something worth sitting with if you run a professional services firm. In the United States, entry-level roles most exposed to AI are now seven times more likely to require skills that used to belong to senior staff: judgement, leadership, the ability to handle a situation nobody wrote a procedure for. PwC's own framing is that AI is removing some of the routine work that historically functioned as an apprenticeship, the repeated, supervised exposure through which junior staff built judgement gradually over years.

That creates a specific problem for firms whose value has always rested on partner and senior-manager judgement. The tasks that used to let junior staff build that judgement through repetition are exactly the tasks AI is now absorbing, and firms need junior staff to display senior-level judgement earlier than that traditional apprenticeship model was ever designed to produce it. PwC's data doesn't tell us how much time any individual has to acquire that judgement. It tells us the skill is now expected earlier, while the mechanism that used to teach it gradually is doing less of that work.

Two trends worth reading together

In the United States, the Bureau of Labor Statistics reported in September 2024 that median tenure across all wage and salary workers had fallen to 3.9 years, the lowest figure since the survey began tracking it in 2002. That figure describes the entire US wage and salary workforce. It isn't specific to knowledge work, professional services or senior experts, and it says nothing directly about the UAE or any other market. It's evidence of a real decline in a large, well-measured labour market, and a useful benchmark for how quickly employment relationships are turning over more broadly.

For firms in the UAE, there's a different structural feature worth naming, even though it doesn't come with a comparable turnover figure attached to it. Foreign nationals make up roughly 85% of the UAE's private sector workforce, a composition that reflects how the market itself was built rather than a recent hiring trend. That fact on its own doesn't establish how long individual employees stay in a given role, and it shouldn't be read as a proxy for higher turnover. What it does mean is that a firm in this market can't default to assuming its expertise sits with people who have spent decades in the same seat, the way a more geographically settled labour market might assume by default.

Neither figure proves that professional services turnover specifically is rising. Together, though, they describe a world where firms can no longer count on long, uninterrupted tenure to pass judgement down the slow way, and that's the reading we take from them, rather than a conclusion either data set reaches on its own.

What actually leaves when someone leaves

Files, precedent databases and matter histories stay behind when an employee resigns. What doesn't automatically stay behind is the reasoning that connected those files to a decision.

Take an auditor deciding whether a discrepancy is material enough to flag. The number itself sits in a spreadsheet anyone can access. The judgement about whether that particular number, for that particular client, in that particular regulatory context, crosses the threshold that matters, usually exists only in the auditor's head, built from years of specific cases that never got written up as a rule. A junior colleague inheriting the file gets the spreadsheet. They don't automatically get the reasoning that would let them make the same call correctly next time.

This is not a criticism of how professional services firms operate. Judgement built this way is exactly what clients pay for, and it resists being reduced to a checklist without losing the nuance that made it valuable. But it does mean the knowledge most worth protecting is also the hardest to protect, because it was never designed to be written down in the first place.

Why documentation alone doesn't solve this

A widely cited 2012 McKinsey Global Institute estimate found that interaction workers spent close to a fifth of their time searching for and gathering information that already existed somewhere inside their organisation. That figure is now over a decade old and describes an information environment that has changed considerably since, but it's still cited constantly for a reason: the underlying problem, that documented information is often difficult to locate even when it exists, hasn't gone away.

That's the easier version of the problem. It concerns information that was written down somewhere and simply needs to be found. The harder version is judgement that was never written down, because nobody thought to ask for it until the person holding it had already left. Building a better search tool or a tidier document repository solves the first problem. It does nothing for the second.

Capturing judgement while it's still in the building

The practical implication is timing. Judgement is realistically only capturable while the person who holds it is still around to be asked specific questions about specific decisions, rather than general questions about their expertise in the abstract. "What do you know about client risk assessment" produces very little. "Why did you flag this particular transaction and not that one" produces something a colleague can actually learn from.

This is one of the areas where AI tools can genuinely help, and where it's worth being precise about the limits. AI can help structure and surface documented knowledge more effectively, and it can prompt more useful, specific capture of reasoning through structured questioning tied to real past decisions rather than generic knowledge-transfer templates. What it cannot do is retroactively recover judgement that was never externalised in any form before someone left. The window for that only exists once, while they're still there to be asked.

What this means for firms right now

Firms whose senior judgement is the actual product they sell are exposed on two sides of the same problem. The apprenticeship model that once transferred that judgement gradually is compressing as AI absorbs the routine work junior staff used to learn from. And in a market like the UAE, where the private sector workforce is overwhelmingly composed of foreign nationals, a firm can't default to assuming that judgement sits with people who will stay in the same seat for a decade while the old, slow model does its work.

Treating knowledge capture as something that happens during an exit interview, after the decision to leave has already been made, addresses almost none of this. The specific reasoning behind a specific past decision is the thing worth capturing, and it has to happen on a rolling basis, while the people who hold it are still making similar decisions every week, rather than as a one-off exercise once someone has already handed in their notice.

AinTools works with professional and financial services firms on turning organisational expertise into a durable, accessible asset. Our industries page sets out how we approach it.

Sources

• PwC, 2026 Global AI Jobs Barometer (official report and press release): pwc.com
• U.S. Bureau of Labor Statistics, "Employee Tenure in 2024" (official news release, 26 September 2024): bls.gov
• UAE private sector workforce composition (approximately 85% expatriate), cited via Envoy Global's summary of UAE government workforce data: envoyglobal.com
• McKinsey Global Institute, "The Social Economy: Unlocking Value and Productivity Through Social Technologies" (2012), cited with its original date and scope: mckinsey.com