As AI takes over more of the work employees once learned from, organizations face a risk that may not appear in their productivity reports until it is too late.
By Dr. Ravinder Tulsiani
Bill Gates recently suggested that society may need to learn something from the Amish.
In an Oct. 22, 2026 interview on The Verge’s Decoder podcast, Gates argued that as artificial intelligence becomes capable of performing more human activities, we may deliberately choose to reserve certain work for people. He gave childcare and food preparation as examples. He also named art and sports.
His point was that technological capability should not automatically determine what we choose to automate. Some activities have value because humans perform them, even when technology offers a faster or more efficient alternative. The Amish example was illustrative, a way of talking about deliberate restraint, not a proposal that everyone should live the same way.
Gates was addressing a much larger societal question about employment, human purpose, and the future of work. But his argument raises an immediate concern for business leaders.
What happens when we automate the very work through which people develop the capabilities an organization will need in the future?
Consider a junior analyst who spends the first few years of their career gathering information, investigating problems, testing assumptions, and preparing recommendations. Much of that work can now be accelerated with AI.
From a productivity perspective, the benefits are obvious. Reports take less time, analysis becomes more accessible, and employees can accomplish more with fewer resources.
But something else happens during those early years. The analyst learns to recognize when the numbers do not make sense, when a conclusion is unsupported, and when the obvious answer is probably wrong. Those experiences gradually develop the judgment expected of a senior professional.
If AI takes over much of that work, how will the next generation develop the same expertise?
We may be improving today’s performance while weakening tomorrow’s capability.
The experience behind expertise
Organizations have always depended on work itself to develop people. Formal learning provides knowledge and practice, but much of professional expertise develops through experience and feedback. Mistakes matter too, as do progressively difficult decisions.
An aspiring manager learns by handling difficult conversations. A financial analyst learns by investigating inconsistencies. A healthcare professional develops judgment through supervised clinical experience. An engineer builds expertise by diagnosing problems whose solutions are not immediately apparent.
These activities produce business value, but they also serve another purpose. They prepare people for greater responsibility.
AI changes that relationship.
When technology performs the analysis, drafts the recommendation, or resolves the problem, the employee may receive a better immediate result without developing the understanding that previously came from doing the work.
This is not an argument that employees must continue performing every task manually. Much of that work is repetitive, and AI can remove genuine obstacles to learning and performance.
The concern is whether we understand which experiences develop essential capabilities before we eliminate them.
Research gives us reason to examine that question carefully. A 2025 study by researchers from Microsoft and Carnegie Mellon University surveyed 319 knowledge workers about their use of generative AI. Higher confidence in AI was associated with less reported critical-thinking effort, while greater confidence in one’s own abilities was associated with more.
The study does not establish that AI causes critical-thinking skills to deteriorate. It does, however, raise an important question about how employees engage with their work when they increasingly trust systems to produce answers.
For employers, the distinction matters. Someone who consistently delivers good results with AI may be highly effective at using the technology. That does not necessarily establish that they understand the underlying problem or can exercise sound judgment when the system fails.
Yet many organizations continue to evaluate AI success primarily through adoption, time savings, and output.
We need to know what is happening to human capability as well.
The leadership pipeline problem
The implications become more serious when we consider succession planning and leadership development.
Most organizations expect employees to progress through increasingly demanding roles. Early assignments build the experience needed for later decisions. Managers learn from operational challenges before assuming broader responsibilities.
Now consider what happens when AI removes significant portions of the work traditionally performed by junior and mid-level employees.
A company may need fewer entry-level analysts while still requiring experienced financial leaders. It may automate routine technical work while continuing to depend on senior engineers who can diagnose complex failures.
Where will that experience come from?
We cannot assume that employees will develop expertise simply because they have access to increasingly sophisticated tools. Nor can we assume that the traditional progression from junior to senior roles will continue to work when the activities within those roles have changed.
We also do not yet have long term evidence that today’s automation of junior work will produce tomorrow’s leadership gap. The risk is real enough to test, and it should be stated as a risk to test, not as a proven outcome.
Organizations will need to examine how expertise is acquired, which experiences remain essential, and whether new development pathways can produce the capabilities required.
This is particularly important because the consequences may remain hidden for years. A reduction in junior roles produces immediate cost savings. A weakened succession pipeline may only become visible when an experienced leader leaves and the organization discovers that nobody is ready to replace them.
By then, rebuilding the missing experience will be difficult.
A different question for the executive team
AI investment decisions generally begin with what technology can accomplish and what the organization stands to gain.
Those questions are necessary. But they are incomplete.
Before automating a significant part of a role, leaders should also ask what employees currently learn by performing that work.
Which experiences develop judgment? Which activities prepare people for greater responsibility? Which capabilities must employees retain independently, even when AI can assist them?
Consider a customer service operation introducing AI to resolve more complex inquiries.
The business may see shorter handling times and lower operating costs. But experienced supervisors often develop their judgment by dealing with unusual cases, recognizing patterns of customer frustration, and navigating situations where established procedures do not provide an adequate answer.
If those experiences disappear, the organization needs another way to develop future supervisors.
That may involve deliberately assigning complex cases to developing employees, creating supervised opportunities to challenge AI recommendations, or assessing employees through unfamiliar situations where they must explain and defend their decisions.
The objective is to preserve the development of expertise, rather than preserve every task that once contributed to it.
This is an organizational design decision with direct consequences for future performance.
Where L&D leadership needs to contribute
For learning and development leaders, this changes the conversation.
Much of L&D’s involvement in AI transformation currently focuses on preparing employees to use new tools, building AI literacy, and helping the workforce adapt to changing roles.
That work remains important. However, it often begins after decisions about automation and work design have already been made.
By that stage, the organization may have eliminated developmental experiences without considering how they will be replaced.
L&D should be involved earlier, alongside business leaders and HR. Technology and workforce planning should be in the room as well.
Its contribution should be to identify the capabilities that future performance depends on, understand how those capabilities develop, and examine whether proposed changes to work will strengthen or undermine them.
This also requires a different approach to measurement.
If employees complete work faster with AI, we should recognize the productivity gain. But we should also test whether they can identify errors, challenge recommendations, explain their reasoning, and respond effectively when circumstances fall outside familiar patterns.
In short, are employees becoming AI-capable, or simply AI-active? A strong output with AI is not, on its own, proof that the person behind it is ready for the next level of responsibility.
We need evidence of what employees can do with AI and what they understand well enough to do without it.
The balance will differ by role. A professional does not need to reproduce every calculation an AI system performs. But someone accountable for a consequential decision must possess enough understanding to recognize when the recommendation is unreliable and know what to do next.
That capability cannot simply be assumed because the final output looks correct.
AI readiness is not AI literacy. It is the organization’s ability to redesign work as fast as AI changes what work can be, including the work that develops people.
The cost of getting this wrong
There is a temptation to treat these concerns as resistance to technological progress. That would be a mistake.
AI offers substantial opportunities to improve performance, expand access to expertise, and remove work that consumes time without contributing much to employee development.
Organizations should pursue those benefits.
But productivity gains should be evaluated alongside the capabilities the organization will need to sustain performance over time.
A business that becomes dependent on AI for routine decisions must still know who can handle exceptional ones, and a company that automates entry-level work must still develop experienced professionals.
These are not hypothetical questions about a distant future. They arise whenever organizations redesign work around AI.
Gates’s proposal to reserve certain activities for humans is ultimately about making deliberate choices rather than allowing technological capability alone to determine how society functions.
Business leaders face a related choice.
They must decide which work should be automated, which human capabilities must be preserved, and how future employees will develop those capabilities when traditional learning experiences disappear.
For L&D leaders, the responsibility is to make those capability consequences visible before decisions are finalized.
Start with one role. List the tasks in that role that build judgment, decide which ones to protect, and test people on an unfamiliar case without AI before you declare them ready.
The question I would put to every executive team is straightforward:
Are we automating work faster than we are redesigning how people become capable of doing it?
If the answer is yes, the productivity gains we celebrate today may be creating a capability problem we will have to solve tomorrow.
