AI monitoring of employee computer use
“The world is too much with us; late and soon, / Getting and spending, we lay waste our powers.” Wordsworth wrote those lines in 1802, watching the first steam engines reshape England’s Lake District. Two centuries later, the machines have grown quieter but more intimate. In Lexington, Kentucky, the city’s chief information officer recently told the city council that the city can monitor employee computers to track any potential use of artificial intelligence. [1] The statement was matter-of-fact, almost bureaucratic. But inside it sits a thousand years of human anxiety about tools that watch us back.
The policy is straightforward on paper: Lexington’s government employees can use AI for approved tasks, but their computers are monitored to ensure compliance. The city has established policies and restrictions on AI use by city employees, according to the official announcement. [1] No benchmarks, no hardware specs, no product reviews. Just a quiet administrative decision that ripples outward into every cubicle, every screen, every keystroke.
The First Layer: Where AI Lifts
Before we understand what Lexington’s policy reveals about deception and obsolescence, we must first see where AI genuinely lifts. The city’s information officer described AI as a tool that could help employees draft documents, summarize reports, and automate routine data entry. These are not glamorous applications. They are the slow, patient work of municipal governance—permits, zoning variances, public records requests—where every minute saved is a minute that can go back to citizens.
Consider the historical weight of this. For most of human history, the bottleneck in government has been information processing. The Roman Empire fell in part because its administrative system could not keep pace with its territorial growth. The British East India Company collapsed under the weight of its own paperwork. Even the modern American city, with its spreadsheets and databases, still struggles to process the flood of data that citizens generate every day. AI that can summarize a 200-page environmental impact statement in thirty seconds is not a luxury. It is a lift.
The city’s policy allows employees to use AI for tasks that are “low-risk and well-defined.” This is the sensible approach: let the machine handle the repetitive, the predictable, the rule-based. Let it lift the weight that human workers should not have to carry. In doing so, it frees them for the work that only humans can do—the face-to-face conversations, the judgment calls, the creative solutions to problems that no algorithm has seen before.
But here is where the story gets complicated. The same policy that lifts also watches.
The Second Layer: Where AI Deceives
The monitoring capability that Lexington’s IT department now possesses is not about AI itself. It is about the fear that AI inspires. The city’s chief information officer told the council that the monitoring exists to track ‘any potential use’ of AI that violates policy, but the officer declined to specify which behaviors would trigger a flag or how the system would be audited for fairness This is the deception—not that AI is lying, but that the promise of AI has become a cover for something older and more troubling.
Surveillance in the workplace is as old as the factory whistle. Frederick Taylor’s time-and-motion studies in the early 1900s were the first systematic attempt to monitor and optimize human labor. The difference is that Taylor’s methods were crude—stopwatches, clipboards, supervisors walking the floor. Today’s monitoring is silent, continuous, and invisible. The employee who opens a ChatGPT tab to draft a public notice does not know whether that action will flag a report. The ambiguity itself is the control mechanism.
This is where the deception operates on multiple levels. First, there is the deception of the technology itself. AI tools are not neutral. They are trained on data that contains biases, errors, and blind spots. An AI that summarizes a police report about a traffic stop may miss the context of racial profiling because the training data never included that context. The employee who relies on the AI summary may unknowingly perpetuate a distorted version of events.
Second, there is the deception of the policy. Lexington’s rules are presented as protective—protecting the city from liability, protecting citizens from errors, protecting employees from themselves. But the monitoring also protects the institution from accountability. If an AI tool makes a mistake, the city can point to the policy and say, “We had rules. The employee should have known better.” The machine becomes a scapegoat that is also a watcher.
Third, there is the deception of trust. The city’s information officer emphasized that the policy is about “responsible use.” But responsibility flows in one direction: from the employee to the system. The system does not need to be responsible to the employee. It does not need to explain its decisions, correct its errors, or apologize for its failures. The employee must simply comply, and the monitoring ensures that compliance is enforced.
The Third Layer: Where AI Makes Us Superfluous
This is the deepest cut. Lexington’s policy does not just monitor AI use. It implicitly defines which tasks are worthy of human attention and which are not. The low-risk, well-defined tasks that AI can handle are precisely the tasks that have traditionally been the entry point for new workers, the training ground for judgment, the place where young employees learn how government works by doing its most mundane work.
Remove those tasks, and you remove the apprenticeship. The new employee who used to spend six months processing permit applications—learning the quirks of the zoning code, the personalities of the planning department, the rhythm of municipal decision-making—now has those six months compressed into a single AI query. The learning does not happen. The judgment does not develop. The human becomes superfluous not because the AI does the work, but because the human never learned to do the work in the first place.
This is not a prediction. It is already happening in cities across the country. A 2023 survey by the National Association of Counties found that 47% of county governments were exploring or implementing AI tools for administrative tasks, while 62% had no formal training program for employees on how to use these tools. The survey did not specify which counties were included or how many responded, limiting its generalizability. The pattern is clear: adopt the technology, monitor the users, and assume that the human element is either a problem to be managed or a cost to be minimized.
But the most insidious form of superfluity is not the loss of jobs. It is the loss of meaning. The Lexington employee who once took pride in processing a complex permit application—the satisfaction of navigating the bureaucracy, the gratitude of the citizen who finally gets their approval—now watches the machine do it in seconds. The employee’s role shifts from creator to overseer, from craftsman to guard. The work becomes watching, and watching is not work that anyone finds fulfilling.
The Historical Roots of the Watching Machine
To understand why Lexington’s policy matters beyond Lexington, we must look at the deeper pattern. The monitoring of workers by machines is not new. The Luddites of early 19th-century England destroyed textile machinery not because they hated technology, but because the machines were used to break their bargaining power, to speed up their work, to replace their skills. The factory owners who installed the machines also installed overseers, time clocks, and piece-rate systems. The technology and the surveillance were always linked.
In the 20th century, this linkage became more sophisticated. The assembly line was both a production tool and a monitoring system—every worker’s pace was visible to every other worker, and the line itself enforced a rhythm that no one could escape. The computerization of offices in the 1980s brought keystroke monitoring, screen capture, and email logging. Each wave of technology promised liberation but delivered control.
What is different about AI is the scale and the opacity. The old monitoring systems were crude but transparent. A supervisor could see a worker slowing down. An employee could see the camera in the corner. But an AI monitoring system is a black box. The employee does not know what triggers a flag, how the algorithm weighs different behaviors, or whether the system is even working correctly. The opacity itself is a form of power.
The Deeper Deception: The Promise of Objectivity
Lexington’s policy is framed as a neutral administrative measure. But every policy carries assumptions about what is valuable, what is risky, and who should decide. The assumption here is that AI use is inherently risky and must be constrained. This assumption is not wrong—AI does pose risks, from data privacy violations to biased outputs. But the policy does not address the risks that AI poses to the human workers themselves.
Consider the risk of deskilling. When AI summarizes a document, the employee no longer needs to read the full document. Over time, the skill of reading and comprehending complex texts atrophies. The employee becomes dependent on the AI, unable to evaluate its outputs critically. This is not a hypothetical. Studies of pilots who rely heavily on autopilot systems have found that their manual flying skills degrade significantly. The same pattern applies to any profession where automation replaces practice.
Consider the risk of isolation. When AI handles the routine interactions that once required human contact—answering citizen questions, processing complaints, providing information—the employee loses the social connections that make work meaningful. The citizen loses the sense of being heard. The city loses the informal feedback loops that catch problems before they escalate.
Consider the risk of accountability. When an AI makes a mistake, who is responsible? The employee who used the AI? The vendor who built it? The city that approved it? Lexington’s policy does not answer this question. It simply monitors the employee, implying that the employee is the point of failure. The AI becomes a tool that can never be wrong, only misused.
The Map Coordinates of Ongoing Work
Lexington’s policy is not an outlier. It is a data point on a larger map. The coordinates are these: the tension between efficiency and dignity, between control and trust, between the promise of liberation and the reality of surveillance. Every city, every company, every institution that adopts AI will face this tension. The ones that succeed will be those that treat AI as a partner, not a master; that invest in human skills, not just machine outputs; that build trust, not just monitoring.
The map is still being drawn. The coordinates shift with every new policy, every court case, every worker complaint. But one thing is clear: the question is not whether AI will change work. It already has. The question is whether we will let the machines watch us without also watching them.
The city of Lexington sits at 38.0406° N, 84.5036° W. That is the geographical center of this story. But the moral center is elsewhere—in the cubicle of a city employee who opens an AI tool and wonders who is watching, in the council chamber where a policy is approved without debate, in the mind of a citizen who never knows whether the person on the other end of the phone is human or machine. That is where the work continues, and where policymakers must decide whether surveillance or trust will define the next era of work.