Two reports landed this week that look like separate stories. One is about hiring. The other is about learning. Read together, they point to the same problem: evaluating real skills has become harder for everyone, on both sides of the desk.

Employers say they can no longer tell what a candidate can actually do. Workers, especially younger ones, say they are spending more time learning than ever and not always getting better at the work.

This is not a talent problem. It is a proof problem.

Key takeaways

  • AI has made the familiar evidence of skill, such as polished applications, completed courses and even interview answers, less reliable as proof of what someone can actually do.
  • In a report released September 30, 2026, Western Governors University found that 60% of U.S. hiring professionals say AI is making it harder to evaluate candidates’ real skills.
  • A Preply survey reported by Fortune on October 1, 2026, found that 71% of Gen Z respondents felt busy learning without improving.
  • Leaders do not need lower standards or more credentials. They need a better proof system: clear standards, and evidence that shows how people think, decide and apply what they know.

What’s happening

On September 30, Western Governors University released its Workforce Decoded report on AI and hiring. It is based on a survey of 3,128 U.S. hiring professionals, conducted by Centiment between June 24 and July 7, 2026. Sixty percent said AI is making it harder to evaluate candidates’ real skills. Nearly one in five cited difficulty confirming whether they were interviewing a human rather than AI. The share of employers struggling to evaluate AI skills specifically doubled, from 16% in 2025 to 32% in 2026.

How are employers responding? Partly by leaning harder on credentials. According to WGU, 73% of employers consider college degrees important, up from 68% in 2025, and 91% view certificates as valuable indicators of skills, up from 86%. At the same time, 42% plan to increase their focus on skills over degrees.

The uncertainty has a cost, and it lands on people trying to get in the door. Among employers who report greater difficulty evaluating skills, 54% say AI has reduced entry-level hiring. Among employers who do not report that difficulty, the figure is 20%. WGU President Scott Pulsipher put it this way:

When employers cannot trust the signals in front of them, opportunity narrows — especially for people trying to get their first foothold in the workforce.

Scott Pulsipher, Western Governors University

A day later, Fortune reported on a Preply survey about AI and learning at work. Preply, a language-learning platform, surveyed more than 5,000 working professionals across nine countries who had tried to develop a career-related skill in the previous year. Among Gen Z respondents, 92% used AI for learning. Yet 57% said they struggled to put what they learned into practice, compared with 40% of Gen X respondents. And 71% of Gen Z respondents said they felt busy learning without improving, compared with 56% of Gen X respondents.

This is not only a Gen Z finding. Across all working adults ages 18 to 55 in the survey, 76% said they spent more time learning than they did two years ago, and 52% said that extra time did not always bring better results. More than half (51%) said their employer paid for some or all of that learning.

Fortune also quoted Iain Smith of Sunny Workplace on what many early-career workers are missing: “They’re not necessarily now sitting next to someone who’s been doing the job for 20 years and learning the shortcuts and the hacks and the little ways of coping with a stressful day.”

As always with generational data, these are one survey’s findings about groups of respondents, not a description of every person in a generation.

What this reveals: the standard did not move, the proof did

In Chapter 11 of The Next Turn, “Where It Breaks,” I describe four pressure points where the collision between the changing definition of work and the changing structure of work shows up first: hiring, performance, retention and communication. This week’s reports sit right on top of the first two.

In my work with leadership teams, when capable people start to feel harder to find, I keep coming back to one distinction from that chapter:

What they are often feeling is not the collapse of standards, but the collapse of the signals they once used as a replacement for standards.

The Next Turn

That is exactly what the WGU data describes. The standard has not changed. Organizations still need people who can think clearly, solve real problems and follow through. What changed is the evidence. A polished cover letter, a clean take-home assignment, a confident interview answer: those used to be reasonable clues about capability. AI can now produce or heavily shape all of them. The clue still looks the same. It just tells you less.

Chapter 13, “Building a Shared Definition of Good Work,” makes the distinction explicit. Standards are what the organization actually needs: quality, reliability, accountability, sound judgment, follow-through. Signals are the visible cues we use to infer those things. “Standards and signals are not the same thing.”

Seen that way, a rising reliance on degrees and certificates makes sense, but it should also make us pause. Credentials carry real information. They are still signals. When a familiar signal weakens, the instinct is to reach for another familiar signal. It feels safer. It does not answer the question the organization is actually asking: can this person do the work, and can they decide well while doing it?

The Preply findings show the same split from the inside. “Busy learning without improving” is what happens when learning activity becomes the signal and capability is assumed to follow. AI makes access to information nearly unlimited. It does not supply the practice, feedback and context that turn information into skill. Much of that used to happen informally, by watching someone more experienced handle real situations. As I argue in Chapter 9, “Invisible Work,” more and more work no longer leaves a visible trail. Learning by proximity is part of what is fading with it.

This is where the two lenses from Chapter 2 help. The definition of work is what people believe work is for and what it should give back. Many younger workers expect a visible return on effort, so learning that does not translate into progress will feel like wasted time, and they will notice quickly. The structure of work is how work actually gets done. AI has made content easy to get and polish easy to produce. Put those together and you get exactly what this week’s numbers show: lots of learning, lots of credentials, and very little confidence on either side about what any of it proves.

The answer is not to lower the bar, and it is not to pile on more requirements. It is to rebuild the proof.

The standards remain. The proof system changes.

The Next Turn

Interestingly, WGU’s president reaches for the same idea, calling for “a better proof system” that includes assessments that verify what a person can do. I agree. And I would add that this is not only higher education’s job. Every team leader decides, every week, what counts as proof.

Your next turn: practical steps for evaluating real skills

  1. Write the standard before the signal. Pick one role you are hiring for. List the three to five standards that actually matter. Then list the evidence you currently use to judge them, and mark which items are signals (degree, certificate, polished writing sample) rather than direct evidence of the standard.
  2. Add one step that shows thinking, not just output. Give candidates a realistic task, tell them whether AI use is allowed, and then ask them to walk you through what the tool did, what they decided, and what they checked. Reasoning is much harder to outsource than a finished product.
  3. Make funded learning prove itself on the job. If your organization pays for courses, pair each one with an application assignment and a short check-in 30 days later. Measure use, not completion.
  4. Rebuild the “sitting next to someone” layer on purpose. Shadowing, narrated decisions, and reviewing real work together give early-career employees the context that AI cannot. It will not happen by accident in a hybrid or AI-heavy workflow.
  5. Do not let uncertainty quietly close the entry-level door. If you are hiring fewer early-career people because you cannot tell who is ready, the fix is a better evaluation, not a smaller pipeline.

Frequently asked questions

Is AI making it harder to evaluate job candidates’ skills?

According to Western Governors University’s September 2026 Workforce Decoded report, 60% of U.S. hiring professionals say it is. The core issue is that common evidence of skill, such as applications, writing samples and take-home work, can now be produced or shaped with AI. Leaders need evaluation methods that reveal how a person thinks and decides, not just what they can deliver.

Should we require more degrees or certificates to make up for it?

Credentials still carry useful information, but they are signals of capability, not the capability itself. Adding more credential requirements can narrow your pipeline without telling you much more about who can do the work. A stronger approach pairs credentials with direct evidence, such as realistic work samples and conversations about reasoning.

How can managers tell if employee training is actually working?

Look for application, not completion. In the Preply survey reported by Fortune, 52% of working adults said more time spent learning did not always bring better results. Ask employees to use what they learned on a real task, then review the work together and talk through the choices they made.

Closing

Leaders are not wrong to feel uncertain right now. The tools they used to recognize capability are less reliable than they were even two years ago. That is a structural change, not a failure of effort or care on anyone’s part.

But the response matters. Organizations that cling to old signals will hire more cautiously, train more expensively, and still not know what they are getting. Organizations that define their standards clearly and build better ways to see them will find capable people others overlook. You can find more on this theme on the Next Turn Leadership blog.

So here is the question I would leave with you: if you removed every credential and course certificate from your evaluation process tomorrow, what would you still trust as proof that someone can do the work?

Sources