Editor’s note: The following is a guest post from Philip Walsh, senior director analyst at Gartner.
Generative AI allows employees to offload far more than routine tasks — workers can now delegate analysis, synthesis, problem solving and even elements of decision-making to AI systems.
The result is often faster work and higher output. But the shift can also create a less visible challenge: Employees may become increasingly disconnected from the reasoning behind the work they are ultimately accountable for.
For CIOs, this is emerging as one of the most important workforce issues of the AI era.
As AI becomes more capable, organizations must ensure employees remain able to evaluate, defend and improve AI-generated outputs.
Workers are increasingly responsible for deciding what thinking should be delegated, how much scrutiny a task requires and whether a result can be understood and defended. Those capabilities become more important as AI assumes a greater share of cognitive work.
Nearly one-third of organizations will identify at least one critical role in which AI use has weakened employees’ ability to exercise independent judgment by 2030, according to Gartner projections.
Workforce readiness in the age of AI will depend less on whether employees can perform every task themselves and more on whether they can exercise sound judgment while working alongside increasingly capable AI systems.
Good outputs don’t mean good judgment
One of the easiest mistakes organizations can make is assuming that acceptable outputs indicate healthy workforce capability. Two employees may produce similarly strong results with AI assistance while demonstrating very different levels of understanding.
One carefully evaluates assumptions, validates evidence and challenges recommendations before acting. The other accepts AI-generated outputs largely because they appear reasonable. Both may produce acceptable work, but only one maintains a strong connection between judgment and outcome.
This distinction often remains invisible until conditions change, assumptions prove incorrect or an important decision needs to be explained and defended. As AI becomes embedded in everyday work, leaders need a more nuanced understanding of human involvement.
Some tasks are routine, low risk and easily validated. In those situations, rapid delegation may be entirely appropriate. Other tasks carry greater uncertainty, higher consequences or require expertise that the organization must preserve. Those situations demand deeper human engagement.
The challenge is determining how much human agency a particular task requires. This is where the idea of “stewardship” enters. When offloading tasks to AI, effective stewardship means the ability to calibrate human involvement based on the stakes, risks and objectives of the work while retaining responsibility for the outcome.
Employees must decide when AI can reliably perform a task, when it should operate within controls and when direct human reasoning is necessary.
Promoting AI stewardship
Critical thinking has long been considered a valuable professional skill, but AI elevates it further into a foundational workforce capability.
As employees rely on AI for more analysis and decision support, they must continue developing the skills required to frame problems, evaluate evidence, challenge assumptions and make defensible decisions.
Stewardship depends on those capabilities because employees must continuously decide what should be delegated, what should be reviewed and what work should remain under direct human control.
The change creates a challenge for CIOs and business leaders. Many of the activities through which employees develop judgment are also the activities AI can now perform on their behalf. Organizations therefore need to identify where critical thinking, comprehension and judgment are formed and ensure those opportunities are not lost in the pursuit of efficiency
Training programs, onboarding experiences and learning pathways should increasingly emphasize explanation and reasoning rather than output. Employees should be expected to validate AI-generated work, explain their conclusions and demonstrate how they arrived at recommendations.
Incentives matter as well. Organizations that solely reward output are encouraging employees to pursue the fastest route to completion. Over time, that can reinforce habitual delegation while weakening independent judgment.
Design new work environments
Judgment is shaped not only by individual capability, but also by the environment in which people work. Organizations cannot expect employees to exercise stewardship consistently if performance expectations, workflows and technology designs continually reward effortless delegation.
CIOs should then focus on creating conditions that encourage thoughtful engagement with AI, where judgment matters most while making efficient delegation easy for routine work.
One effective approach is what Gartner describes as “useful friction.”
Useful friction introduces moments of reflection, challenge and accountability at points where judgment materially affects outcomes. It can take many forms, including peer review, assumption testing, structured challenge processes or accountability mechanisms that require employees to defend key recommendations.
The value of these practices extends beyond the review itself. The expectation of review changes behavior earlier in the process: Employees who know their reasoning may be challenged are more likely to examine assumptions, validate evidence and engage critically with AI-generated outputs.
Technology design can reinforce these habits. Organizations building internal AI applications have an opportunity to design systems that elicit judgment. AI tools can ask clarifying questions, surface assumptions, present competing interpretations or challenge conclusions before producing recommendations.
AI systems will continue to become more capable. Organizations will still need employees who understand when to delegate, when to supervise and when direct human reasoning remains necessary. And CIOs that focus exclusively on productivity gains may eventually discover that critical capabilities have eroded beneath improving output metrics.
