I’ve been watching the headlines about government workforce reductions for months now, and something keeps nagging at me. On the surface it looks like progress toward leaner operations. Fewer desks, fewer salaries, fewer layers of paperwork. Yet the deeper I dig, the more it feels like we are not shrinking government at all. We are simply swapping human judgment for code that never sleeps, never questions, and never refuses an order.
When Efficiency Starts Replacing Restraint
The numbers are hard to ignore. Hundreds of thousands of federal positions have vanished through freezes, early retirements, and quiet reductions. At the same time agencies receive clear instructions to eliminate, optimize, and automate wherever possible. The official story frames this as modernization. I see something more unsettling. We are building a system that needs fewer people to carry out its will and even fewer people willing or able to push back when that will collides with fairness or the law.
Human beings hesitate. They ask awkward questions. They sometimes refuse. Machines do none of those things. For an administration tired of delays and institutional friction, that quality is not a bug. It is the main selling point. And that, frankly, is where the danger begins.
Fewer Workers Does Not Mean Less Power
People still cling to the old idea that smaller staff equals smaller government. That equation no longer holds. A government that employs fewer humans can still grow far more powerful if those remaining systems can surveil, classify, investigate, and decide at machine speed. The new digital gateway now being built is a perfect example. Citizens will eventually interact with agencies through a single AI-powered portal that handles forms, records, and transactions with almost no human touch.
On paper the benefits sound attractive. Backlogs shrink. Jargon gets translated into plain language. Tedious tasks disappear. I have no problem with tools that free people from repetitive work. The trouble starts when those tools stop assisting and start deciding. When the software becomes the investigator, the gatekeeper, the analyst, and eventually the enforcer, the entire relationship between citizen and state changes.
We are moving from a society governed by laws and deliberate process to one increasingly governed by software that executes without pause.
Your tax return, your disability claim, your immigration status, your travel patterns, your associations, even your threat profile can all be processed automatically. In the most extreme scenarios some military planners already discuss, the same logic could decide who qualifies as a target. That is not efficiency. That is the removal of every deliberate delay the Constitution was designed to create.
Why Constitutional Friction Exists in the First Place
Government is not a private company. Its purpose is not simply to move faster or spend less. It holds powers no corporation possesses: the ability to tax, search, arrest, imprison, seize property, and in rare cases take life. Because those powers are so absolute, speed cannot be the highest value. Liberty has to be.
Due process is inefficient. Warrants take time. Trials are slow. Appeals drag on. Human review introduces delay and disagreement. Checks and balances are spectacularly inefficient by design. Those obstacles exist for a reason. They force government to justify its actions before it acts. Artificial intelligence promises to cut through exactly those obstacles. That is wonderful when you need an appointment. It becomes terrifying when the decision involves your freedom or your safety.
For generations manpower itself acted as a natural limit. There were only so many investigators, so many analysts, so many hours in a day. AI removes that limit. One algorithm can examine millions of records. One system can cross-reference data that would have taken teams of people years to assemble. The surveillance state no longer needs enough eyes. It needs enough computing power. That shift changes everything.
The Accountability Problem Nobody Wants to Face
Critics of the administrative state have long complained about unelected officials holding too much power. Replacing the official with an algorithm does not solve the problem. In many ways it makes the problem worse. At least a human bureaucrat has a name. Someone can be questioned. A supervisor can intervene. An employee can refuse an unlawful order. A whistleblower can speak up. Responsibility, however imperfect, still exists.
What happens when the answer becomes “the computer decided”? Who do you cross-examine? Who explains the reasoning? Who takes the fall for a catastrophic mistake? And what happens when the code itself belongs to a private company whose training data and internal logic remain hidden behind trade secrets? Government power stays government power, but accountability quietly migrates into the private sector where public oversight is weakest.
We already see early versions of this shift. Agencies experiment with AI tools that review records requests and recommend what should be withheld. Human reviewers still sign off for now, yet the direction of travel is clear. Tasks that once required judgment about public access are becoming candidates for automation. The same pattern appears in border systems, eligibility determinations, and risk scoring. Each step looks modest. Together they form a new operating system for the state.
From Paperwork to Life-and-Death Decisions
The stakes rise dramatically once artificial intelligence moves from forms to force. Recent disputes between the Pentagon and major AI developers centered on exactly this question. One company insisted on hard limits against mass domestic surveillance and fully autonomous weapons. Officials pushed back, arguing that such restrictions interfered with operational needs. The disagreement itself should worry anyone who cares about limits on power.
The issue was never whether AI can assist military personnel. The real question is whether a human being must always remain responsible for certain irreversible decisions. When nations resist international efforts to require human review of AI-generated targeting recommendations, the answer becomes clearer. Speed and scale are winning the argument over caution.
I keep returning to one reported incident. Missiles struck a school that had once been part of a military compound but had operated openly as a school for years. Outdated intelligence still listed the site as a target. That flawed data fed into an AI-assisted targeting system. Officials relied heavily on the technology while an accelerated process compressed hours of work into minutes. Civilian-harm teams had been reduced. The machine did not independently decide to bomb a school. Humans built a system that prioritized speed and scale, then deferred too much judgment to it. That distinction matters. It is also precisely the danger.
- First the software handles paperwork
- Then it recommends decisions
- Then it makes decisions
- Eventually it decides who lives and who dies
At every stage the temptation grows to hand over a little more judgment because the machine is faster, cheaper, more consistent, and easier to blame when things go wrong. Technology does not erase bias, political pressure, or misconduct. It can automate them, amplify them, and hide them inside a black box that few people understand and even fewer can challenge.
Self-Policing Is Not a Safeguard
Assurances that the industry will police itself should comfort no one. Voluntary accords, internal audits, and corporate ethics statements sound reassuring in press releases. History offers little reason to trust power when it promises to restrain itself. The entire structure of constitutional government rests on the opposite assumption. People are not angels. Neither are the corporations building these systems or the officials deploying them.
James Madison put it plainly centuries ago: if men were angels, no government would be necessary. The Founders divided power because they understood that good intentions and voluntary promises are not enough. Algorithms do not change that oldest lesson. If anything they make the lesson more urgent. Power still expands until it meets resistance. Removing the human beings who can provide that resistance does not eliminate bureaucracy. It eliminates the brakes.
The Long-Term Infrastructure Problem
Current leaders will not remain in office forever. The databases, automated decision systems, surveillance tools, and public-private partnerships being built right now will outlast any single administration. Once those capabilities become embedded in the machinery of government they become extremely difficult to dismantle. Future officials of every political stripe will inherit them and, almost certainly, expand them.
That is why the debate cannot be reduced to job counts or cost savings. The larger question is what happens when fewer and fewer human beings stand between raw government power and the people subject to it. What happens when there is no human face behind the decision, no human hesitation, no human capacity for mercy, and no one left who can simply say this is wrong?
The greatest risk is not that artificial intelligence will seize control of the state. The greater risk is that officials will willingly surrender more and more control because machines are obedient, tireless, and convenient to blame. That is how bureaucracy without bureaucrats becomes authoritarianism without accountability.
A government can employ fewer people, close offices, eliminate forms, and still watch more citizens, investigate more lives, process more cases, and exercise more power than any earlier generation of officials could have imagined. That is not smaller government. That is government without human limits.
What Meaningful Limits Would Actually Look Like
If we accept that artificial intelligence will play a growing role in public administration, the conversation has to shift from whether to how. Certain principles seem non-negotiable to me. High-stakes decisions that affect liberty or life must retain meaningful human review. The reasoning behind automated recommendations needs to be explainable in language ordinary people and courts can understand. Private code that shapes public power cannot remain completely opaque. And the same constitutional protections that once constrained human officials must constrain the systems that replace them.
None of this is anti-technology. I use advanced tools every day and appreciate how much tedious work they eliminate. The issue is not the existence of the technology. The issue is the systematic removal of friction, judgment, and accountability from the exercise of coercive power. Efficiency is a means. It cannot become the end.
Perhaps the most interesting aspect is how quietly this transformation is occurring. Most citizens experience only the surface benefits: faster responses, simpler forms, fewer trips to government offices. The deeper architecture of power is changing beneath those conveniences. By the time the full implications become obvious, the new systems may already be locked in place.
A Personal Note on Why This Matters
I am not a Luddite. I have watched technology improve ordinary life in countless ways. What keeps me up at night is the combination of two trends: the rapid reduction of human discretion inside government and the simultaneous expansion of technical capacity to monitor and classify entire populations. Either trend alone would deserve careful scrutiny. Together they create a new kind of state, one that can act faster and more comprehensively than any previous version while offering fewer points of resistance.
The Constitution was written for a world in which power moved slowly and required many human hands. We are building a world in which power can move at machine speed with almost no human hands at all. That mismatch is not a minor technical issue. It is a fundamental challenge to the idea of limited government itself.
We still have time to insist on real safeguards. But the window is closing as more systems come online and more human roles disappear. Once the infrastructure of automated governance becomes the default, rebuilding the human brakes will be extraordinarily difficult. The choice is still ours. The longer we treat this purely as an efficiency project, the more likely we are to discover too late that we have engineered away the very restraints that made constitutional government possible in the first place.
In the end the question is simple. Do we want a government that processes us, or a government that remains answerable to us? The tools now being deployed will push strongly in the first direction unless we deliberately design them otherwise. That design work cannot be left to the same institutions and companies that benefit from removing the friction. It requires public attention, legislative clarity, and a renewed commitment to the old idea that power, no matter how sophisticated its machinery, still needs limits that only human beings can enforce.
