When AI becomes the manager: Why young workers want more explicit instructions

Artificial intelligence is changing how work gets done, but it may also be changing what workers expect from managers. New findings supplied by the AI platform Use.AI suggest that younger employees who regularly use AI tools are developing a different idea of what constitutes a clear assignment.

According to the Use.AI findings, 62 percent of regular AI users aged 18 to 28 prefer managers to provide highly specific instructions, including clearly defined steps, detailed parameters and an indication of what the finished result should look like. Among regular AI users aged 40 and over, 41 percent expressed the same preference. This creates a 21 percentage point difference between age groups within a population that already uses AI regularly. The findings indicate a divide that may extend beyond the adoption of technology. They raise questions about how different generations approach task definition, professional knowledge, autonomy and judgement in an increasingly AI-assisted workplace.

The influence of prompting

Anyone who regularly uses generative AI quickly learns that the quality of an output is influenced by the clarity of the request. AI users frequently specify context, constraints, format and the intended outcome before asking a system to complete a task.

For younger employees, this experience may be affecting expectations of human managers. A broad instruction that an experienced manager regards as an opportunity for initiative may be interpreted by an AI-accustomed worker as an incomplete prompt. This does not necessarily mean that younger workers lack independence. It may instead indicate that they have become accustomed to systems in which the requester defines the task explicitly and the recipient produces an output within stated parameters. The question for organisations is whether this preference leads to better execution or gradually reduces opportunities to practise independent problem-solving.

Producing an answer is not the same as understanding it

The most significant Use.AI finding may concern reasoning. The research indicates that 44 percent of regular AI users aged 18 to 28 have struggled to explain the reasoning behind work completed with AI assistance. This issue is supported by wider research into AI-assisted knowledge work. A study from Microsoft Research and Carnegie Mellon University surveyed 319 knowledge workers and examined 936 examples of generative AI use in workplace tasks. The researchers found that higher confidence in generative AI was associated with less critical thinking, while greater confidence in a worker’s own ability was associated with more critical thinking.

The study did not conclude that AI simply eliminates critical thinking. Instead, it found that AI can shift critical activity towards verifying information, integrating generated responses and overseeing the completed task. This distinction matters. Creating a polished report, presentation or analysis does not necessarily demonstrate that the person submitting it understands the evidence, assumptions and decisions behind it. The risk is particularly important in regulated or high-consequence sectors. In pharmaceuticals, healthcare, engineering, finance and law, work must frequently be traceable and defensible. An employee may be required to explain not only what was concluded, but why the conclusion is valid.

AI and the changing purpose of memory

Use.AI also reports that 57 percent of regular AI users aged 18 to 28 are less likely to deliberately memorise information when they know it can be retrieved or generated using AI. The shift resembles an expansion of cognitive offloading. Search engines reduced the need to retain certain factual information because it could be retrieved quickly. Generative AI goes further by summarising sources, comparing arguments and constructing responses. This may bring genuine benefits. Workers can devote less time to routine recall and more time to interpretation, creativity or decision-making. Yet foundational knowledge remains important because it enables people to identify when an AI-generated answer is inaccurate, incomplete or based on a weak assumption.

UNESCO’s guidance on generative AI in education and research argues for a human-centred approach that develops human capacity while protecting agency. UNESCO’s guidance also stresses ethical validation, meaningful use and the long-term implications of generative AI for education and research.  Although education and employment are different settings, the underlying issue is similar. If people rely on AI before acquiring sufficient subject knowledge, they may become good at generating plausible material without being well equipped to evaluate it.

The Use.AI findings also point to an emotional shift. Some 53 percent of younger regular AI users said routine workplace tasks feel disproportionately frustrating when they know AI could complete the work in seconds. This reaction is understandable, since repetitive administration can appear especially inefficient when a readily available system can draft, classify or summarise information almost immediately. Yet not every routine task exists solely to create an output. Some tasks help employees recognise patterns, understand processes, develop professional discipline or notice anomalies that automated systems may overlook.

The OECD’s work on artificial intelligence and skills highlights the importance of skills in securing the benefits of AI and identifies workforce development as a significant policy issue. Its broader Skills in the AI Age report examines how AI is changing labour markets and skill requirements, while stressing the need to equip workers for the transition. The managerial challenge is therefore not merely deciding which tasks can be automated. It is determining which activities develop capabilities that remain important when automation is available.

This leads to the central management problem raised by the Use.AI research. Managers must decide whether a task is primarily about producing an output or whether completing the work is also intended to develop judgement. If the purpose is a standardised output, detailed instructions and AI assistance may be entirely appropriate. Clear parameters can improve consistency, reduce rework and help employees understand the expected standard.

If the purpose is professional development, excessive prescription may be counterproductive. Employees need some opportunity to frame problems, weigh evidence, test alternatives and defend their decisions. The two aims can also be combined. A manager might permit AI assistance but require the employee to identify the sources used, describe key assumptions, explain rejected alternatives and defend the final recommendation. Such an approach changes the manager’s role. Instead of assessing only the finished product, the manager evaluates whether the employee understands the work sufficiently to take responsibility for it.

The findings also support the case for stronger AI literacy. Workers need to understand both the capabilities and limitations of generative systems. The OECD’s AI and skills programme notes that technological developments can change how people learn and that AI is altering tasks and skill requirements across jobs. It also identifies leadership, communication and teamwork as important transversal skills within AI-related work.

Research involving young people similarly shows that AI is already strongly associated with learning and information seeking. A Harvard Graduate School of Education overview reported findings from research into teenagers’ and young adults’ perspectives on generative AI, including its use for obtaining information and brainstorming. This makes AI literacy more than technical proficiency. A skilled user should be able to formulate an effective request, test the response, locate original evidence, recognise uncertainty and explain the final reasoning in their own words.

The Use.AI results do not prove that AI is making younger employees less capable. Nor do they imply that older workers are inherently more independent. The supplied findings describe preferences and self-reported experiences among regular AI users, not fixed characteristics of entire generations. Nevertheless, they raise an important organisational question. If younger workers increasingly expect managers to provide instructions resembling well-constructed AI prompts, management practices may need to become more deliberate.

The answer is not to make every instruction prescriptive. Instead, managers can state the purpose of the task, distinguish mandatory requirements from areas of discretion and explain whether the assignment is being used to assess output, judgement or both. Where judgement matters, employees should be asked to explain their reasoning. Where efficiency is the priority, organisations should consider whether routine work can be automated safely rather than insisting on manual activity without a clear purpose.