AI at Work: When AI Efficiency Comes at the Cost of Engagement

Quantifying Employee Engagement

At Terryberry, we quantify employee engagement through six connected indicators: purpose, leadership, belonging, equity, empowerment, and well-being. This model gives organizations a practical way to break down, categorize, and assess the health of its employees.    

Efficiency is easy to count. Leaders can measure the time required to complete a task. They can calculate the labor needed to support it.   

Engagement is less obvious. It lives in whether employees believe their work matters and if they trust the people leading them. Those measurements rarely appear next to cost savings on a financial dashboard. They usually become visible later, after morale drops or strong employees begin looking elsewhere.

 

Purpose 

Purpose may be the clearest place to see the risk. Work can be valuable to a company without feeling meaningful to the employee doing it. Training an AI model may help companies like Meta achieve its strategic goals, but that does not mean the assignment feels purposeful to an experienced engineer whose broader abilities are no longer being utilized.

A 2026 study published in Scientific Reports found that passive reliance on AI reduced feelings of ownership over work by nearly 20 percent. Meaningfulness and self-efficacy declined by almost 10 percent. Interestingly, some "efficacy and meaningfulness" remained impacted even after employees returned to manual work.

The same research found a different outcome when people actively collaborated with the technology. Those employees did not experience the same loss of ownership. AI was not inherently stripping meaning from the work. The damage appeared when the person’s contribution became passive.   

This distinction should matter to every organization introducing AI. Is the technology helping employees apply their abilities at a higher level? Or is it slowly moving them out of the creative process?   

 

Leadership  

Leadership guides how employees answer those questions. People need context during major organizational change. They also need a credible explanation of where they fit.   

TheOECD surveyed more than 6,000 managers across six countries about algorithmic management. Many believed the technology improved their decisions. Yet 64 percent reported at least one concern about its trustworthiness. In the United States, that figure reached 82 percent. What happened when employees had influence over implementation? The negative effects on autonomy and motivation largely disappeared. 

This is where leadership has a significant impact. Declaring  

that AI is now part of everyone’s job does not help employees understand their future. Tracking usage does not create confidence in the strategy. Leaders need to bring employees into the change and give them some influence over how their work evolves.   

 

Belonging  

Belonging is another indicator we measure that is seeing a significant impact. Work is a social environment, even when interactions happen through remote meetings, chat, and email. Employees build relationships by solving problems together and asking one another for help.   

A series of studies published in the Journal of Applied Psychology found that employees who interacted more frequently with AI experienced greater loneliness. That loneliness followed employees even after they clocked out, as employees reported increased anxiety and declined sleeping habits. 

There was a telling counterpoint, however. Some employees responded toward this isolation by becoming more helpful toward coworkers. Researchers suggest these individuals are reacting as a sort of counterculture to corporate AI initiatives.     

Meta employees have described limited contact with their reassigned teams. Some were also reluctant to speak openly in meetings captured by AI note-taking systems. It is difficult to create belonging when employees are uncertain who is listening or how their words may be used. 

  

Equity  

Equity is another indicator that surfaces with additional questions. In Meta's case, there was a flattening of roles as experienced leaders were suddenly being asked to do the same tasks as employees who had been there longer or who had fewer scholarly accolades.   

Not only did this affect an employee's sense of purpose as described above, but it also played into a sense of equity as employees may begin to wonder why they are being paid less when completing the same type of work.    

In addition, while automated measurements can at first appear neutral when it comes to productivity, it also creates a narrow image of the ideal employee. A system built around constant activity may disadvantage someone who takes protected leave. It could also penalize an employee who needs an accommodation.   

That concern sits at the center of the Meta lawsuit. The employees allege that measures such as productivity and AI usage placed workers with medical conditions at a disadvantage. Meta says the workforce decisions were made by people.   

The wider research shows why the question deserves scrutiny. In four experiments involving more than 4,400 participants, workers who used AI were often perceived as less competent. They were also viewed as less motivated.  

Employees now face pressure to demonstrate that they are using AI while knowing others may judge them for relying on it. Organizations must examine who benefits from their expectations and who is quietly penalized by them.  

 

Empowerment  

When we discuss empowerment, we see it as more than providing access to technology. Employees need the autonomy to question an AI response. They should also be able to decide when the tool is appropriate.   

Research from Microsoft and Carnegie Mellon found that greater confidence in AI was associated with less critical thinking. Employees increasingly moved from creating work to verifying machine-generated output. That shift can improve speed, but it also changes the employee’s role. Requirements change that relationship even further.    

Once a company tracks AI activity or connects usage to performance, the technology becomes a compliance exercise. Employees are no longer choosing how AI can help them. Instead, they are simply proving it was used at all.   

 

Well-Being

Finally, these pressures accumulate into well-being. The American Psychological Association found that workers concerned about AI making their duties obsolete were much more likely to feel tense during the workday. They were also more likely to report lower motivation.  

Now consider Meta’s employees. Thousands of people lost their jobs while others were moved into unfamiliar work. Those who remained were asked to improve technology that could reduce future staffing needs. Anxiety in that environment is not a refusal to innovate. It is a reasonable response to uncertainty.   

The research does not suggest that AI inevitably damages employee culture. Used well, it can help less experienced employees build capability and remove remedial tasks that prevent meaningful outcomes.    

The conditions surrounding AI adoption and the way leadership communicates that vision will ultimately impact an employee's experience. They need transparency about how the technology will be used. More importantly, they need a meaningful role in deciding how AI will impact their work.   

This is the real tension between efficiency and engagement. AI may help organizations produce more with fewer resources. If those gains weaken purpose or trust, however, the organization is exchanging one form of performance for another.