A new survey has given leaders a useful name for something many of them have been sensing: AI experience debt. It is the capability an organization quietly gives up when AI takes over the tasks that used to teach people how to do the work.

Most of the AI conversation is about what gets faster. This one is about what stops getting built.

The work still gets done. The practice reps disappear. And the judgment those reps used to produce does not show up on its own.

Key takeaways

  • AI experience debt is the future cost of people losing chances to learn by doing when AI takes over the tasks that used to build their skills.
  • In Dayforce’s new survey of nearly 7,000 people, 65% said their people are losing the ability to perform some tasks AI now does for them.
  • This is a structure-of-work problem: AI changes how work flows, and the old, accidental path to judgment goes with it.
  • Leaders now have to design how judgment gets built on purpose, instead of assuming experience will deliver it.

What’s happening

This week, Dayforce released new research on workforce fragility and AI. HRD reported on October 27 that the report, based on a survey of nearly 7,000 respondents across six countries, warns that employers may be losing skills and expertise because AI tools are taking over tasks that once helped people learn. The report’s own definition is worth reading slowly: “We call this experience debt: the future capability cost created when people lose opportunities to learn by doing.”

The numbers HRD reported are striking:

  • 65% say their people are losing the ability to perform some tasks that AI now does for them.
  • 68% say organizations are losing important skills or expertise without fully realizing it.
  • 73% warn that these capabilities are harder to rebuild.
  • 54% say important knowledge is concentrating among fewer people.
  • 50% say they are expected to develop new skills while keeping the same workload, and only 27% say they have enough time during work hours to do it.

HR Dive covered the same findings on October 28, noting that almost 9 in 10 respondents were concerned about AI’s impact on workforce development, that 40% expect their employer to simply hand them more work if AI saves them time, and that Dayforce researchers concluded “leaner workforces have less room for disruption.”

The same week, HR Dive interviewed Toshiba CHRO Jason Desentz (October 27) about early-career talent. His view runs in a different direction from the usual worry. He said new graduates often arrive without business AI skills, in part because many universities discourage AI use, and that Toshiba is responding by giving interns Copilot licenses and offering an AI readiness course that covers business use cases and building agents. His summary: “AI is really not the story. Adaptability is the story.”

Put those together and you get a real tension. Entry-level people need to use AI more to be ready for the work. And the more AI does the work, the fewer chances people get to learn the work underneath it.

Both are true. That is exactly why this is a leadership question.

What AI experience debt reveals about the structure of work

In The Next Turn, I describe two turns happening at once. The first is the definition of work: what people believe work is for and what it should give them. The second is the structure of work: how work actually flows, who or what does each part, and how value gets produced. Experience debt sits squarely in the second turn.

For a long time, organizations built judgment by accident. A new analyst built the spreadsheet by hand and learned where the numbers lie. A new coordinator wrote the first draft of every update and learned what the audience cared about. A new supervisor handled the routine questions and learned which ones were not routine at all. Nobody called that a development program. It was just the work.

AI removes a lot of that work. That is the point of it. But when the structure changes, the side effects of the old structure change too. The learning that used to come free with the task does not come free anymore.

In Chapter 7, “The Collapse of Effort,” I argue that AI does not make effort disappear. It relocates it:

AI does not eliminate effort. It often moves it upstream or changes its form.

The Next Turn

The upstream work is framing the problem, deciding what good looks like, checking what the tool produced, and catching what it got wrong. That is where the learning now has to happen. The problem is that upstream work is quieter. It leaves less of a trail. A new employee can skip it entirely and still hand in something that looks finished.

Why it matters for leaders: judgment doesn’t build itself

Chapter 8, “The Rise of Judgment,” is the heart of this. As output gets easier to produce, judgment becomes the thing that separates people:

What increasingly separates people is not whether they can produce a first version. It is whether they can decide what that version is worth.

The Next Turn

Here is the trap. Polished AI output can make someone look capable earlier than their decision quality actually justifies. That is not a character flaw in early-career employees, and it is not unique to any generation. Anyone handed a tool that produces confident output can be pulled into trusting it too soon. In the book, I put it plainly: “That is not an employee problem. That is a leadership design problem.”

Experience debt is what that design problem looks like at scale. If the only path to judgment was doing the routine work, and the routine work is gone, then the path is gone. Unless someone builds a new one.

Notice also what the Dayforce data says about time. Half of respondents are expected to build new skills on the same workload, and only about a quarter say they have the time. That is a familiar pattern: asking people to grow while the system gives them no room to. Readiness is a structure question, not a willingness question.

This is where I would push back gently on framing this as “AI versus skills.” Desentz is right that adaptability matters and that people need real fluency with these tools. The Dayforce researchers are right that something valuable is leaking out. The answer is not less AI. It is being deliberate about which parts of the work are for learning, not just for producing.

Your next turn

Here are five moves you can make with your team this week:

  • Map the old learning path. Pick one role. List three tasks that used to teach new people the job. Ask whether AI now does any of them, and if so, where that learning happens today.
  • Name the learning reps. Decide which tasks early-career people should still do by hand, or do first and then compare against AI, because the point of the task is the learning, not the output.
  • Ask about the decision, not just the deliverable. In one-on-ones, try: “What did the tool do here, and what did you decide?” and “What did you check, and what did you trust?” Those questions make the upstream work visible.
  • Protect real time for development. If people are expected to build new skills, put it in the workload. Decide explicitly where some of the AI-saved time goes, before it gets absorbed by default.
  • Spread the knowledge on purpose. If expertise is concentrating in a few people, pair them with newer colleagues on judgment-heavy work, and have them narrate how they decide.

Frequently asked questions

What is AI experience debt?

It is a term from Dayforce’s new research for the future capability cost created when people lose opportunities to learn by doing, because AI now handles tasks that used to build their skills. The work still gets done, but the experience that used to come with it does not.

Should we limit AI use for entry-level employees to protect their skills?

Not as a blanket rule. Early-career people need business AI fluency to be ready for the work. A better approach is to be specific: identify the tasks where doing it yourself is the learning, and pair AI use with explicit expectations to explain choices, check output and make the final call.

How can managers tell if someone has good judgment when AI produces the output?

Evaluate the decisions, not just the deliverable. Ask people to talk through how they framed the problem, what they verified, what they rejected and why. Reasoning that holds up under a few honest questions is a far better signal than polish.

The work is still human

For most of working history, experience was something that happened to people. You did the work long enough and judgment showed up.

That arrangement is quietly ending. AI can carry more of the production. It cannot carry the formation. Leaders now have to design the path to judgment on purpose, or accept that their organization will be fast today and thin tomorrow.

If you want more on how the structure of work is shifting, you’ll find related pieces on the Next Turn blog.

So here is the question I would sit with this week: in your organization, where are people still learning to make the call, and how would you know if that learning had stopped?

Sources