Before I begin, I would like to take you to an experiment at the world's second-largest mobile phone factory in southern China. For years, the production lines there have been laid out close together and in the open. The purpose of this layout is simple: the manager should see every line and, on spotting a good method, spread it to the other lines. We see the same logic in most production facilities.
Ethan Bernstein of Harvard Business School put a curtain around one line and observed it for 5 months. The line behind the curtain turned out to perform roughly 10-15% better than the 28 open lines.
The reason: the workers on the open lines already knew faster methods. When the manager walked by, they dropped these methods and returned to standard procedure; once the manager moved away, they went back to their own methods. What the manager saw was a performance of production. When the curtain made that performance unnecessary, the line switched to its own best method.
Bernstein called this the transparency paradox. In short: making people more visible can lower performance, because it pushes them to hide their own methods.
This experiment was carried out in 2012. Today, if the owner of a 150-person company is researching digital tracking software because they can no longer see their team the way they used to, we can think of them as being inside a similar approach.
The setup built in a factory in 2012 now lives inside technology
The mechanism Bernstein found has 3 steps. The employee knows they are being watched. They guess what the watcher approves of. They display the approved behavior instead of their own method.
These 3 steps are fully in place in office work today. Monitoring is now done through signals: active hours on the computer, keyboard and mouse movement, app usage time, the green light in the messaging tool, meeting attendance, email response time. The approved behavior is for these signals to come out high.
The result is the same as in the factory. Employees take actions to keep the screen active, leave their status on green, attend meetings that give them visibility and generate unnecessary message traffic. Microsoft calls this cluster of behaviors productivity theater.
This is no longer a problem of a few companies. According to Gartner's research, the share of large employers monitoring their employees with tracking tools has doubled since the start of the pandemic, reaching 60%.
Microsoft's 2022 Work Trend Index data also showed the tension building around efficiency and productivity. 85% of leaders said they find it hard to be confident their teams are productive. In the same period, 87% of employees said they were productive. Microsoft called this gap productivity paranoia.
How the loop works: when managers lose the visibility link with their remote and hybrid teams, distrust grows; demand for measurement and monitoring follows and, most importantly, when measurement drops down to behavior, employees behave according to what is measured. The resulting data reassures the manager, but has real performance actually been measured?
So does monitoring work? The shared answer of 94 studies
The answer to this question can in fact draw on many studies. Ravid and colleagues conducted a meta-analysis: they combined all studies on the same topic into a single pool and calculated the common result. The pool contains 94 independent studies and 23,461 employees. We can call it the experience and accumulated knowledge of many companies.
3 findings emerged. There is no evidence that electronic performance monitoring increases employee performance. The presence of monitoring, whatever its form, is associated with employee stress. And companies that monitor openly and with low intrusiveness achieved clearly better results in employee attitudes.
I believe the third finding is more impactful than the first two. Monitoring itself did not work as a performance tool. How it is done, however, directly shapes how employees view the company. Rather than "should we monitor?", the important question is "what, for what purpose, and how openly?"
In practice the difference is this: a company that installs cameras for workplace safety and tells everyone so in writing, and a company that quietly uses the same cameras in performance evaluations, make the same investment and get opposite results.
The real bill lands on culture
The picture so far is an efficiency debate. Where we need to look, however, is this:
The study by Thiel and colleagues published in the Journal of Management shows that constant monitoring weakens employees' capacity to regulate their own behavior. A person controlled from the outside hands the responsibility for right behavior over to the system. The result is the opposite of what is expected: among monitored employees, rule-breaking increases instead of decreasing. The authors also summarized the finding in Harvard Business Review.
This finding is critical for family businesses. These companies' strongest asset is often not a corporate system but the knowledge that "this is how things are done here". That knowledge stands on the responsibility employees have internalized. Intense monitoring weakens exactly this muscle.
The rule is this: when measurement looks at outcomes, the employee remains the owner of the work. When measurement drops down to behavior, ownership passes to the system. This is where culture gets lost.
A company that runs a values program on one side and builds a behavior-tracking infrastructure on the other cancels out one investment with the other.
What really increases performance?
Another study conducted in the same period shows the other side of the coin.
Brynjolfsson, Li and Raymond studied the gradual rollout of an AI assistant at an organization with 5,172 customer support agents. The number of issues resolved per hour rose by 15% on average.
The real finding is in the distribution. Most of the gain occurred among inexperienced and low-performing employees. In the most experienced group, the effect was close to zero.
The mechanism: the system captures the tacit knowledge of the best employees. Tacit knowledge is knowledge no one has written down, which stays in the master's hands and head. Every company has it, and it leaves when that person leaves. As the assistant channeled this knowledge to new employees, learning time shortened. The researchers also report that customer satisfaction rose and attrition fell.
Put the two studies side by side. Monitoring moves knowledge from the employee to the manager and leaves it there. The AI assistant moves knowledge from the best employee to the others. The first is the logic of control, the second the logic of diffusion. This is the approach that creates the performance difference.
Is technology alone enough to produce output?
The classic example from economic history is this: when factories bought the electric motor, productivity did not rise immediately. The increase came after factory layout, workflow and the foreman's role were redesigned. Decades passed in between.
Brynjolfsson, Rock and Syverson call this delay the productivity J-curve. General-purpose technologies such as AI do not turn into measurable productivity without complementary investments that require rebuilding processes, roles and human capabilities. Because these investments do not appear on the balance sheet, productivity is underestimated in the early period.
From a management perspective: the return on the budget you set aside for software will not materialize without rebuilding role definitions, competencies, the learning system and manager behavior. The technology bill alone will not deliver transformation.
4 rules of practice for growing companies
1. Measure the outcome, leave the method to the employee. Bernstein's curtain did exactly this. Sales figures, delivery time, error rate and customer satisfaction are measured. How many hours someone sits at the screen is not. The moment the method starts being measured, employees hide their method.
2. State clearly what you collect. What data is collected, for what purpose it is used, who sees it, how long it is kept. One page is enough. This is where the meta-analysis's only positive finding pays off. If employees can access their own data, the effect is even stronger.
3. Give AI to new and inexperienced teams first. Common practice is to give tools to senior staff first. The research argues the opposite. An application that shortens new employees' ramp-up time plays an important role in both productivity and attrition.
4. Bring the master's knowledge into the system. Turn how your top performer works into writing, training, checklists and job descriptions. This produces a higher return than tracking software and keeps the knowledge in the company when that person leaves.
Together, the four will also change culture: responsibility stays with the employee, knowledge sharing is rewarded, and newcomers ramp up faster.
If you would like to see where your organization's technology and culture stand today, the IC Consulting Check-Up portal is a good starting point: consulting-ic.com/check-up
Sources in this issue
Bernstein, E. S. (2012). The Transparency Paradox: A Role for Privacy in Organizational Learning and Operational Control. Administrative Science Quarterly, 57(2), 181-216.
Ravid, D. M., White, J. C., Tomczak, D. L., Miles, A. F., & Behrend, T. S. (2023). A meta-analysis of the effects of electronic performance monitoring on work outcomes. Personnel Psychology, 76(1), 5-40.
Thiel, C. E., Bonner, J., Bush, J. T., Welsh, D. T., & Garud, N. (2023). Stripped of Agency: The Paradoxical Effect of Employee Monitoring on Deviance. Journal of Management, 49(2), 709-740.
Brynjolfsson, E., Li, D., & Raymond, L. (2025). Generative AI at Work. The Quarterly Journal of Economics, 140(2), 889-942.
Brynjolfsson, E., Rock, D., & Syverson, C. (2021). The Productivity J-Curve: How Intangibles Complement General Purpose Technologies. American Economic Journal: Macroeconomics, 13(1), 333-372. NBER working paper.
Microsoft (2022). Hybrid Work Is Just Work. Are We Doing It Wrong? Work Trend Index Pulse Report.
Gartner. The Right Way to Monitor Your Employee Productivity.
For those who want to go deeper: Kellogg, K. C., Valentine, M. A., & Christin, A. (2020). Algorithms at Work: The New Contested Terrain of Control. Academy of Management Annals, 14(1), 366-410. A comprehensive review that breaks down 6 mechanisms of algorithmic control.
