Pressure to prove AI adoption can hide workplace problems

Pressure to prove AI adoption can weaken the feedback leaders need

Leaders who press employees to demonstrate AI adoption can make deployment harder to evaluate. When employees feel expected to report success, their accounts of tool use may hide workflow difficulties, security concerns or a need for support. More reported use can therefore give leaders a less complete picture. To judge whether AI is useful, secure and adequately supported, leaders need employees to explain how it changes their daily work.

Pressure to show results can hide workflow and security problems

Getting that explanation is difficult when leaders have limited visibility into daily work. Jacob Appleton, Visier’s director of information security, told CIO Dive that chief information officers, or CIOs, and chief information security officers, or CISOs, often miss what happens at lower levels of their organizations and underestimate how much AI changes employees’ work. Appleton brings a commercial perspective: Visier has a business stake in how employers manage technology and their workforce. His warning concerns the gap between leaders encouraging AI use and employees making the practical changes it requires.

Security problems can emerge when pressure to increase use outpaces preparation. Appleton warned that leadership demands for AI adoption can increase risk when the organization lacks an adequate security foundation. Employees may be doing what their organization expects while it remains unprepared for the resulting changes in how work gets done. A positive adoption report can hide that mismatch.

Preparation also has to keep pace on both sides of the deployment. “If an organization’s rushing ahead of where their employees are, or pushing their employees to rush ahead of where leadership is, then you have these risks building up in your organization without necessarily having that visibility,” Appleton said. Employees need to understand how AI fits their responsibilities, while leaders need to understand the security implications of the uses they encourage. If either side moves ahead of the other, risks can build outside leadership’s view.

Pressure to demonstrate results makes those differences harder to report. Derler describes demands at each organizational level: CIOs need to demonstrate return on investment, managers need to increase their teams’ AI use, and employees need to find applications that make sense for their work. Each demand asks a different question, so a manager seeking evidence of adoption may get an answer before an employee has established whether an application is worthwhile. Derler also describes severe workflow disruption and differences in employees’ experiences, which makes it essential to allow uneven reports of progress.

Without that room, employees may give managers the answer they seem to want. “If you don’t allow people to see that, they will just give you a blanket answer: ‘Oh yeah, I’m using AI for everything, and everything is better,’” Derler says, illustrating a hypothetical employee response. Derler links this lack of openness to wasted time and unsuccessful organizational adoption. When every experience is presented as a success, managers lose the information they need to decide which applications to continue, change or support.

Real AI benefits coexist with pressure and job concerns

Distorted feedback matters because employees also report genuine benefits. They describe faster task completion, help with collaboration and new career opportunities. These benefits give leaders a reason to explore further deployment, but understanding where they occur requires examining the work behind the adoption report. An employee who completes a task faster gives managers a specific outcome to investigate.

But that investigation must also include the consequences of faster work. Employees describe pressure to produce more, concerns about the security of sensitive information, fear of losing their jobs and expectations that AI could reduce advancement opportunities. Experiences vary, so a broad question about successful AI use can combine several separate judgments. Someone may value help with a task while still worrying about productivity expectations, information security or career prospects.

Job-security findings make these differences more concrete. A recent survey conducted jointly by CNBC and SurveyMonkey found that about one-quarter of employees expressed concern that AI would cause them to lose their jobs. The survey also found that only employees using AI daily felt positively about its effect on their job security. Frequency of use therefore matters when interpreting employee sentiment.

Yet the relationship between daily use and confidence allows different explanations. Frequent use may increase confidence, while employees who already feel confident may be more willing to use AI every day. The sentiment measures employees’ expectations about job security; actual employment protection is a separate question. Managers can take that confidence seriously while checking whether each employee has a useful application and adequate support.

Individual experience also helps leaders interpret productivity reports. Faster work can be real while other employees struggle with disruption or feel pressure to give an encouraging answer. Managers therefore need to identify which tasks improve, which become harder and which concerns remain unresolved. Genuine successes offer useful examples to investigate, while candid accounts of difficulty show where deployment needs to change.

Self-training shows how employees are learning

Understanding what needs to change also means finding out how employees acquire AI skills. In a recent survey by TrustedTech, a commercial technology provider with a stake in business technology spending, 41% of workers said they had learned to use AI tools themselves, while 15% credited employer-sponsored training. Self-directed learning was the most commonly reported source of AI training. Employees’ own learning efforts therefore matter when assessing readiness.

Independent learning can coexist with useful employer support, so managers need to ask how each employee learned and where help is still needed. Someone who teaches themselves may also receive assistance from their employer, and employees learning on their own can face different levels of pressure to demonstrate adoption. The practical question is whether they have the knowledge and support for the work being encouraged. That includes knowing how to handle sensitive information as well as how to operate the tool.

Employees’ requests for greater confidence extend beyond operating skills. They identify training and upskilling, greater transparency from leadership, career advancement opportunities and clearer communication about AI’s effects on employees. These requests connect the ability to do the work with the consequences of becoming more capable with AI. An employee can understand a tool and still need answers about what its adoption means for their future.

Because these needs differ, a demand for more use can miss the support an employee actually requires. Workflow disruption may call for a change in the task, a skills gap may call for training, and an advancement concern may require a clearer conversation about career opportunities. Managers need to investigate each issue directly rather than infer readiness from tool use or self-training. They can then make more specific decisions about what support to provide.

Evaluate use cases individually and make problems safe to report

Assessing each application separately turns those differences into management decisions. Derler advises managers to examine AI use cases, specific tasks or activities where AI is applied, individually because employees are likely to experience them differently. A manager can ask whether an application helps complete the task, supports collaboration or disrupts the workflow. Establishing that someone uses AI begins the assessment; understanding its effect on the work determines what to do next.

To get that account, employees must be able to describe limitations and failures openly. Managers should ask where AI creates extra difficulty, where it fits poorly and where employees have security concerns, while making clear that an unsuccessful application is useful information. Employees then have room to explain mixed results without presenting their whole experience as positive. Those details help managers decide whether to continue the application, change the workflow or provide more support.

Support decisions also need to account for leadership’s preparedness. Alongside employees’ training needs, managers should check whether security arrangements are adequate for the uses being encouraged and whether leaders understand the resulting changes in work. Conversations should include advancement concerns and employment effects because employees identify both as relevant to their confidence. For each application, the assessment should connect evidence of use with the task being performed, the changes it creates and the issues that still require attention.

Main highlights

• Protect candid feedback: Pressure to demonstrate AI adoption can hide workflow problems and security risks. CIOs and CISOs need employees to report difficulties openly so they can assess whether deployment is outpacing preparation.
• Examine benefits and job concerns together: Employees report faster work alongside pressure to produce more and fears about job security. Managers need to assess task improvements alongside changes in workload expectations and career prospects.
• Check readiness beyond self-training: In a TrustedTech survey, 41% of workers credited self-directed learning for their AI skills, while 15% credited employer-sponsored training. Managers need to identify gaps in task-specific skills, sensitive-information handling and communication about AI’s employment effects.
• Evaluate each AI application separately: Managers need to examine which tasks improve, which workflows become harder and which concerns remain unresolved. Those findings inform decisions to continue an application, change the workflow or provide targeted support.