Mateusz Przepiórkowski is CEO of Appsfactory International, focused on AI leadership, digital transformation and future-ready organizationsgettyOver the past few weeks, I found myself in a series of conversations that all pointed in the same direction, even if they started from very different places. At both the NAB Show and Google Cloud Next in Las Vegas, nearly every discussion eventually circled back to how dramatically AI is compressing the time from concept to execution.Ideas that once required weeks of coordination, development and iteration can now be turned into something tangible in a matter of hours. Code is generated almost instantly, content appears on demand, and interfaces can be designed, tested and refined with surprisingly little effort.The excitement around this is entirely justified. But what happens when we no longer need to learn the things that once built our expertise? That nagging thought has stayed with me longer than the demos or announcements.Shaping JudgmentIt brought me back to a very different period of computing. In the early 1990s, developers in the demoscene pushed platforms like the Commodore 64, Amiga and Atari ST to their limits, often relying on the assembler to extract every bit of performance from the hardware. There were no high-level tools, no abstraction layers and no shortcuts. If something worked, it was because you understood exactly how the system behaved, down to memory constraints, processor cycles and timing.It was slow, sometimes frustrating and often unforgiving. But it forced a level of understanding that's difficult to replicate today. Those constraints did more than limit what could be built. They shaped how people thought. You learned not just how to produce an outcome but why it worked—and where it would eventually break.Today, we're operating in a completely different environment. Much of that friction has been removed and most of it deliberately. Tools are designed to abstract complexity, accelerate delivery and make output easier to generate.That's progress. At the same time, it changes something less visible: how skills are formed.Skills don't just produce results. Over time, they shape judgment.And judgment rarely comes from successful outputs alone. It develops through repetition, through mistakes and through understanding why something failed instead of simply moving past it.Reducing Effort = Reducing LearningAI, by design, optimizes for outcomes. It's extremely effective at producing results quickly. What it doesn't replicate is the experience that leads to deeper understanding.It is now possible to generate code without fully understanding the architecture behind it, to create content without deep knowledge of the subject or to communicate across languages without learning them. Each of these capabilities reduces effort.Less obvious is that they also reduce the learning that used to come with that effort. Recent research reinforces how quickly this transition is taking place. The 2025 Microsoft Work Trend Index points to a growing “capacity gap” in organizations, where more than half of leaders say productivity needs to increase while most employees report already being stretched. AI is increasingly seen as the mechanism to close that gap, with many organizations turning to digital labor and AI agents to extend what teams can deliver.At the same time, the expectation remains that humans will provide judgment, creativity and decision making. Those capabilities are still seen as distinctly human, even as more of the execution layer becomes automated.According to the Stanford AI Index Report, generative AI adoption and investment continue to accelerate globally, while productivity gains are becoming increasingly measurable in structured tasks.​ But although AI delivers clear productivity gains in structured tasks, its effectiveness is far less consistent in areas that require deeper reasoning. There are also early signals suggesting that heavy reliance on AI may reduce opportunities for people to develop underlying skills over time.​This is where the tension becomes harder to ignore. On one hand, organizations are gaining speed, efficiency and scale. On the other, some of the processes that traditionally built expertise are being shortened—or bypassed entirely.The risk isn't that people suddenly become less capable. It's that capability becomes more dependent on tools and less grounded in understanding.Why Human Capability Still Matters​In stable environments, that distinction may not matter much. Systems behave as expected, and outcomes are delivered.In moments of disruption, it matters a great deal.Organizations don't operate on output alone. They rely on people who can interpret ambiguous situations, question assumptions and respond when things don't go according to plan. Those capabilities aren't built through efficiency. They're built through engagement with complexity.This doesn't mean stepping back from AI. The benefits are too significant. But it does mean recognizing that efficiency and capability aren't the same thing, and that optimizing for one doesn't automatically preserve the other.Leaders are now in a position to dramatically increase output, but they also influence how the next generation develops its skills. The way work is structured today will shape the level of judgment available in organizations in the years to come.Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?