63% of workers fear AI will make work less human

Employees fear AI will reduce the human value of work

63% of workers say AI will make the workplace feel less human, according to Resume Now’s AI and Workplace Humanity Report. That concern is broader than job loss. The same report found that 57% expect AI to reduce human skills, while 43% believe it will devalue human work. Only 16% think AI will make the workplace more human.

These numbers point to a management problem. Companies are adopting AI to increase output, automate repetitive tasks, and make decisions faster. Employees are asking a different question: What happens to their role when software starts doing work that previously required human effort?

The answer will shape workplace culture. Poor implementation can reduce opportunities for employees to think through problems, exchange ideas, and build expertise. It can also make employees question the value of their own contribution. Resume Now found that 20% of workers expect AI to create a “cold, machine-driven environment.” That perception matters even when the underlying technology is improving productivity.

Executives should therefore measure more than hours saved or tasks automated. They need to identify which human capabilities each AI deployment changes. If AI generates a first draft, who develops the judgment needed to evaluate it? If it handles analysis, where do employees learn to challenge assumptions? If it replaces routine interaction between colleagues, what happens to knowledge transfer?

The goal should not be to protect every existing process from automation. That would sacrifice useful productivity gains. The stronger approach is to automate work where AI has a clear advantage while deliberately retaining human judgment, accountability, learning, and collaboration where they add value.

This distinction is important for workforce planning. AI can change what a job contains without eliminating the need for the job. Leaders should define which tasks move to AI, which remain with people, and which require both. That gives employees a clearer basis for developing the skills the organization will need next.

The constraint is not whether companies can deploy more AI. They can. The harder task is ensuring productivity gains do not come at the cost of the human capabilities the business still depends on. Culture should therefore be part of AI architecture and governance from the start.

Clear AI plans and employee involvement will determine adoption

Employees need to know what an AI system will do, why the company is introducing it, and how it will affect their work. Without those answers, uncertainty fills the gap. That is especially important when employees are exposed to conflicting claims about AI-driven job losses, productivity gains, and changing skill requirements.

Kaelyn Lowmaster, director analyst in the Gartner HR Practice, argues that leaders “must be clear and transparent with their AI strategy and principles.” She also recommends giving employees formal channels to raise concerns, ask questions, and propose AI use cases.

This employee input has practical value. The people performing a process often know where time is lost, where errors occur, and where automation could create new problems. Employees who already use emerging AI tools can also identify useful applications that central technology teams may overlook. Lowmaster notes that this source of knowledge will become more important as AI-native employees enter the workforce.

Communication alone, however, is insufficient. Employees need operational clarity. Leaders should explain which tasks are likely to change, what remains under human control, what new skills will be required, and what the company currently knows about future roles. Executives should avoid guarantees they cannot support. A credible near- to mid-term plan is more useful than broad assurances about long-term job security.

Gartner research cited by Lowmaster supports this approach. She says clarity about employees’ current value and how their roles will change in the near to medium term drives employees to use AI more than any other form of organizational support. The implication for executives is clear: adoption depends on role design as much as access to technology.

Companies should also turn employee participation into a working governance process. Feedback channels can reveal where AI outputs are unreliable, where policies are unclear, and where employees are avoiding useful tools because they do not understand the rules. They can also surface effective AI use cases that can be tested and expanded.

The strongest AI strategy therefore connects technology deployment with workforce planning. Define the business objective. Specify what AI will and will not do. Assign human accountability. Explain how roles will change. Then give employees a structured way to influence implementation.

This approach does not remove uncertainty. No executive can reliably predict every effect of AI on future jobs. It does something more useful: it gives employees enough clarity to act today while giving management better information for the next deployment decision.

AI adoption must preserve human collaboration

AI can increase individual output while reducing interaction between employees. That trade-off deserves executive attention. If employees use AI for brainstorming, reviewing work, and solving routine problems, they have fewer reasons to consult colleagues. Efficiency can improve while knowledge sharing declines.

Kaelyn Lowmaster, director analyst in the Gartner HR Practice, identifies this as a specific risk. Employees who become overly reliant on AI to “brainstorm and review their work” may collaborate less with coworkers. That matters because collaboration does more than complete a task. It transfers knowledge, exposes weak assumptions, develops judgment, and helps less experienced employees learn how decisions are made.

Executives should not respond by restricting useful AI tools. They should redesign work so collaboration remains deliberate. Megan Slabinski, district president of technology talent solutions at Robert Half, recommends “dedicated mentorship time, team-based projects, and in-person or hybrid touchpoints.” These practices can preserve employee connections even as more individual tasks become automated.

Mentorship is particularly important. Junior employees traditionally develop expertise by performing basic work, receiving feedback, and observing experienced colleagues. AI can now complete some of that basic work quickly. Companies therefore need to consider whether automation is removing activities that previously helped employees develop professional judgment. If it is, leaders need a more explicit process for training and review.

The same principle applies to team design. Not every AI-assisted task needs group involvement. Requiring unnecessary meetings would simply replace one inefficiency with another. Executives should instead identify work where human exchange has clear value: complex decisions, review of important outputs, creative development, high-risk cases, and work that depends on knowledge from several functions.

Leaders should also monitor how AI affects performance expectations. Lowmaster warns that if AI increases individual efficiency, employees may face “unsustainable pressure to hit elevated, AI-driven targets for speed or output.” Productivity gains should not automatically translate into proportional increases in workload. Some of the capacity created by AI can support deeper review, collaboration, learning, or higher-quality decisions.

The objective is straightforward. Use AI where it reduces low-value effort, but preserve human interaction where collaboration improves quality, develops skills, or controls risk. Companies that make that distinction explicitly can gain efficiency without weakening the working relationships their performance still depends on.

Automation does not remove human accountability

AI can complete a large share of some workflows. It cannot assume organizational responsibility for the result. That distinction should guide how executives design AI-enabled operations.

Frank Antezana, CEO of iTech AG, says AI may complete “80% or 90% of a workflow,” while the final layer still requires people to validate results, make decisions, and assume responsibility. He also notes that AI is accelerating information processing, task automation, and decision-making without eliminating the need for human judgment and oversight.

The percentage should not be treated as a universal benchmark. Different workflows have different levels of complexity and risk. The important point is that the amount of work AI performs and the amount of responsibility it carries are separate questions. An automated system can produce most of an output while a human remains accountable for whether that output is accurate, appropriate, and safe to use.

This becomes critical when AI produces plausible but incorrect information. Generative AI systems can make factual errors, omit relevant context, or produce answers unsupported by reliable evidence. Human review therefore needs to be designed around risk rather than added as a generic final step.

For a low-impact internal draft, limited review may be enough. Decisions involving customers, employees, finances, security, legal obligations, or other material business risks require stronger controls. Leaders should define who approves the output, what that person must verify, and what happens when the AI result cannot be trusted.

Human oversight must also be substantive. A process does not become safer merely because an employee clicks an approval button. Reviewers need enough subject knowledge, time, authority, and access to source information to challenge the system. Otherwise, human review becomes procedural rather than an effective control.

This has implications for AI governance. Every important AI-assisted process should have clear ownership. Management should know which system generated an output, what information was used, where validation occurs, and who holds final decision authority. Higher-risk uses may also require documented review, testing, monitoring, and escalation procedures appropriate to the organization’s regulatory and operational obligations.

AI can still create substantial value under these constraints. Automation should remove repetitive effort and accelerate analysis. Human expertise should concentrate on exceptions, verification, consequential decisions, and accountability. The executive task is to define that boundary before scaling the technology, rather than discovering it after an AI failure.

Poor AI use can create more work than it removes

AI does not create productivity by default. If employees lack the skills to use it well, they can generate low-quality material faster. Coworkers must then verify, correct, or recreate that work. The organization has automated production without reducing total effort.

This problem is often referred to as AI “workslop”: low-quality AI-generated output passed to colleagues who must make it usable. The cost is easy to miss because the employee producing the original work may appear more productive. The additional effort appears elsewhere in the workflow.

This changes how executives should measure AI returns. Time saved by the initial user is an incomplete metric. Leaders also need to account for review time, error rates, rework, rejected outputs, and downstream delays. An AI tool that saves two hours during creation but creates three hours of verification and correction has reduced productivity.

Training is therefore an operational requirement. Employees need to understand what AI does well, where it is unreliable, how to provide effective instructions, and when independent verification is required. They also need clear rules for sensitive data, confidential information, and high-risk decisions.

Kaelyn Lowmaster, director analyst in the Gartner HR Practice, notes that “overreliance on AI tools” can lead to “cases of poor employee judgment or low-quality output.” More importantly, she says one of the “biggest barriers” identified by Gartner research is a “lack of trust in the accuracy of AI-generated output.”

That lack of trust can erase efficiency gains. If employees assume every AI output may be wrong, they may spend substantial time checking it. If they trust AI too readily, errors may reach customers, managers, or business systems. Neither outcome is desirable. Companies need validation requirements that reflect the risk of each use case.

Accountability is another constraint. Employees should not be able to transfer responsibility for poor work to an AI system. If a person submits an AI-assisted analysis, recommendation, or document, the organization needs to define who is responsible for checking its quality. Clear ownership reduces the chance that unreliable outputs move downstream without proper review.

Executives should judge AI deployments by end-to-end performance. Measure net time saved, output quality, rework, error rates, and the amount of human verification required. Where results are weak, the answer may be better training, a different workflow, stronger controls, or abandoning an unsuitable AI use case. More AI use is not the objective. Better business performance is.

AI productivity gains can create unsustainable performance targets

AI can help an employee complete some tasks faster. Management can then make a costly mistake: treating every productivity gain as permanent capacity for additional work.

Kaelyn Lowmaster of the Gartner HR Practice warns that when AI “boosts individual employees’ efficiency,” workers may face “unsustainable pressure to hit elevated, AI-driven targets for speed or output.” This is a performance-management issue as much as an AI issue.

The key constraint is that AI does not accelerate every part of a job equally. It may produce a draft in seconds, but an employee still needs time to verify facts, resolve exceptions, coordinate with colleagues, exercise judgment, and approve the result. Raising targets according to the fastest automated step can create unrealistic expectations for the complete workflow.

Executives should also distinguish higher volume from higher value. AI can help employees produce more documents, analyses, code, or customer responses. That does not automatically mean the organization has created proportionally more useful output. Quality, accuracy, customer outcomes, and business impact remain more meaningful measures than raw production.

Aggressive targets can also encourage poor AI practices. Employees under pressure to maximize output may reduce verification, rely too heavily on generated content, or pass low-quality results to colleagues. This connects performance management directly to the “workslop” problem. A company can create incentives that increase visible output while also increasing hidden rework.

The better approach is to establish an evidence-based baseline after AI deployment. Measure where time is actually saved and where new work appears. Some tasks will become substantially faster. Others will gain little. New requirements for review, governance, and quality control may consume part of the capacity AI creates.

Leaders can then decide where the remaining capacity creates the most value. In some roles, higher output targets will make sense. In others, employees can spend more time on customers, complex decisions, collaboration, skill development, or quality improvement. The allocation should reflect business priorities rather than an assumption that every minute saved must become another unit of production.

AI should improve the economics and quality of work. If productivity targets rise faster than the technology’s proven net benefit, the organization risks converting an efficiency gain into greater workload, weaker quality, and employee resistance. Executives should scale expectations only after measuring the complete workflow.

Position AI as support for employees

How leaders define AI’s role will influence whether employees use it willingly. Megan Slabinski, district president of technology talent solutions at Robert Half, says companies that position AI “as more of a support tool, rather than a replacement, will likely see stronger employee interest.”

That position needs to be supported by operating decisions. Employees will notice quickly if management describes AI as a support tool while using it mainly to reduce headcount. Leaders should explain which tasks they intend to automate, which responsibilities remain human, and how jobs are expected to change. The message needs to match the implementation.

The distinction between tasks and jobs is important. AI can automate parts of a role without removing the need for the person performing it. It can process information, produce drafts, summarize material, and handle repetitive work. Employees still provide context, validate outputs, resolve exceptions, collaborate across functions, make consequential decisions, and take responsibility for results.

Slabinski emphasizes this continued human role. “Businesses will always need professionals who can apply the technology and collaborate across teams,” she says. For executives, this means AI strategy should include workforce design. The question is not only which processes AI can automate. Management must determine how employee skills and responsibilities should change as automation expands.

This approach also addresses the concerns identified in Resume Now’s AI and Workplace Humanity Report. The report found that 63% of workers think AI will make workplaces feel less human, 57% believe it will reduce human skills, and 43% expect it to devalue human work. Only 16% believe AI will make workplaces more human. Positioning AI as support will have limited value unless employees can see evidence that their judgment, expertise, and development remain important.

Executives should therefore identify the benefit AI is expected to deliver at the employee level. In a strong use case, AI can remove repetitive work, reduce search and drafting time, or help employees process information faster. Management can then redirect some of that capacity toward work where human contribution has greater value, including customer interaction, complex problem-solving, collaboration, and final decisions.

Training should reinforce the same model. Employees need more than instructions for operating an AI tool. They need to understand when to use it, when to challenge its output, what must be independently verified, and where human approval remains mandatory. AI capability and professional judgment need to develop together.

The objective is not employee acceptance for its own sake. It is productive adoption. Companies need employees who use AI where it creates measurable value, question it where reliability is weak, and retain responsibility for their work. Positioning AI as support works when organizational design, performance measures, training, and management decisions all reinforce that principle.

The bottom line

63% of workers already expect AI to make the workplace feel less human. Executives should treat that number as an implementation signal. AI adoption can deliver real productivity gains, but those gains will be harder to sustain if employees lose trust in the technology, its outputs, or management’s intentions.

The priority is not maximum automation. It is better business performance. Automate tasks where AI creates measurable value. Keep human accountability where judgment matters. Train employees to verify outputs. Protect collaboration where it develops knowledge and improves decisions. Set performance targets based on net productivity gains.

Communication must be equally concrete. Explain which tasks will change, which decisions remain human, and how roles are likely to evolve. Do not promise that jobs will never change. Give employees the information they need to understand their current value and prepare for what comes next.

Companies that get this right will not need to choose between AI productivity and a human workplace. They can use AI to remove low-value effort while giving employees more capacity for judgment, collaboration, and higher-value work. That is the standard executives should use to judge whether an AI deployment is actually working.