AI could redefine work in public administration.
Few structures have proven to be as resistant to change as bureaucracy. Built on printed forms, single windows, and linear processes, bureaucracy has long been synonymous with rigidity and inefficiency. But that is beginning to change. Artificial intelligence has emerged as a viable solution to administrative bottlenecks, both in the public and private sectors.
Automating bureaucratic tasks is no longer a futuristic idea; it’s a growing practice. Governments, including Singapore’s, already use AI-powered chatbots to respond to thousands of administrative requests in seconds. In the private sector, companies like Coca-Cola are not only automating contract and invoice processing, but they are also generating brand creative assets using AI—within minutes and at global scale. Employees have been elegantly “reassigned” to higher-value tasks, because as history has shown, human labor doesn’t disappear—it just relocates.
This is neither anecdotal nor temporary. It’s a clear example of the Ricardo Effect at work in the labor market—a concept developed by economist David Ricardo and later expanded by Friedrich Hayek, which explains how rising wages or cheaper capital lead firms to replace labor with machines. As the cost of automating routine tasks drops, firms and institutions substitute low-productivity human labor with capital—in this case, algorithms. This lowers costs, improves quality, and frees up human resources. As people shift toward roles where their marginal productivity is higher—supervision, analysis, and process design, for example—economic output grows. According to the McKinsey Global Institute, this process could boost global GDP by 0.8% to 1.4% annually, provided that displaced workers are efficiently redeployed. Generative AI alone has the potential to add $2.6 to $4.4 trillion per year to the global economy. Combined with other automation technologies, it could contribute up to 3.4 percentage points to global GDP growth by 2040. It’s hard to imagine a more powerful (and less expensive) economic policy than allowing these tools to flourish, especially in bureaucracy.
Market economies don’t exist to preserve jobs as they are, but to reallocate resources to where they create the most value. That’s exactly what’s happening. AI is replacing tasks, not people. Repetitive administrative functions—copying data, tracking forms, forwarding emails, reviewing files—can now be performed by machines more quickly, more accurately, and at a lower cost. The result isn’t unemployment. Instead, it’s functional reorganization. And it creates jobs we can’t even imagine today. Just as in past waves of innovation—agricultural mechanization, office digitization, or media digitalization—workers are being moved toward areas further from the final consumer. There, they help design, build, maintain, and supervise the very tools that now do their old jobs. AI frees time and talent, but it also requires adaptability. The more flexible the labor market, the less pain is felt in the early stages of change.
Not everyone welcomes this shift. Many government officials, realizing that AI could finally streamline processes, reduce headcounts, and eliminate bottlenecks, have panicked. So instead of riding the wave, they’ve thrown up barriers. In Brussels, the AI Act claims to protect “fundamental rights,” but in reality, it imposes traceability requirements, pre-deployment checks, and disproportionate penalties on any AI system deemed “high-risk”—a category that, ironically, includes many of the most useful administrative applications.
The result is predictable: small administrations, startups, and local organizations cannot afford to take on the regulatory cost of implementing these solutions. Only big tech will be able to comply. Access to automation, therefore, is restricted to those who already dominate the market. In the name of ethics, efficiency is frozen, projects are stalled, and adoption is discouraged.
Meanwhile, countries like the United States, India, and Estonia are freely experimenting with AI to optimize their systems—from San José training civil servants to use ChatGPT, to India deploying AI chatbots for citizen support, and Estonia integrating AI across government through its national “Kratt” strategy. Much of Europe, by contrast, entrenches itself behind the “precautionary principle,” mistaking caution for inertia. Innovation cannot thrive under the constant threat of penalties. Markets don’t need to be told what tools to use. They need the space to try them. AI shouldn’t be treated as a threat, but as a technology to be discovered through use. Real-world application, trial and error—that’s what will show us whether a chatbot improves public service or speeds up licensing. Regulating AI as though it were a pharmaceutical drug denies its evolving, contextual nature.
Perhaps the real question isn’t whether AI will take the bureaucrat’s job, but whether we need to keep paying bureaucrats to do tasks a machine can perform 24 hours a day at a fraction of the cost. If development continues, it’s perfectly reasonable to imagine routine administrative functions disappearing from the public payroll and reappearing in the private sector, where those same workers might actually create value. AI doesn’t undermine public employment—it relieves it, redefines it, and gives it meaning. Because there’s a difference between serving the citizen and stamping forms in triplicate. This reallocation of labor is not a modern anomaly. It’s the same process we saw during the Industrial Revolution. As Henry Hazlitt once wrote in Economics in One Lesson:
There is also an absolute sense in which machines may be said to have enormously increased the number of jobs. The population of the world today is three times as great as in the middle of the eighteenth century, before the Industrial Revolution was well under way. Machines may be said to have given birth to this increased population; for without the machines, the world would not have been able to support it. Two out of every three of us, therefore, may be said to owe not only our jobs but our very lives to machines.
The global population has more than tripled since the publication of Economics in One Lesson (1946) precisely because progress allowed us to sustain more people, with more specialization, more free time, and more opportunities. Machines don’t impoverish; they expand the reach of human intelligence. That’s exactly what AI can do for bureaucracy: replace rigidity with agility, paperwork with decisions, queues with instant service.
This isn’t about eliminating people; it’s about giving them better roles. The wealthiest societies aren’t those that cling to obsolete jobs, but those that create new ones where human talent matters more. If we allow AI to transform our administrations, we might also transform how we experience work, time, and progress. Unfortunately, bureaucracy will likely survive. The difference is that from now on, it can be managed by someone who doesn’t take coffee breaks or ask for days off.
