AI Is Not Just a Technology Revolution. It Is a Test of Government.

For most of the last several decades, we have tended to think about technology as something the private sector creates and government eventually regulates. Artificial intelligence makes that distinction increasingly inadequate. AI will certainly change businesses, but it will also change employment, taxation, education, healthcare, financial markets, public administration, infrastructure, cybersecurity and even the way citizens interact with public institutions.

Whether we like it or not, governments themselves will become some of the largest users of AI. That makes the challenge before us much larger than deciding how to regulate a new class of technology. AI is going to test whether our shared institutions are capable of managing a period of extraordinary economic and social change while also using the same technology to make themselves more capable.

I am optimistic about what AI can do. From the perspectives of government, finance, and technology, the potential is extraordinary. But optimism is not the same thing as complacency. The question is no longer whether AI will transform our institutions and our economy. It is whether we will deliberately shape the transition or simply react to its consequences after they arrive.

One of the first mistakes government will likely make is treating AI primarily as an information technology issue. Much of the public-sector conversation today understandably revolves around questions such as whether employees should be allowed to use generative AI, which vendors are acceptable, what information can be entered into these systems, and what cybersecurity and privacy controls should apply. These are necessary questions, but they are much too narrow.

AI is simultaneously a workforce issue, a financial issue, an economic-development issue, an infrastructure issue, a cybersecurity issue, a procurement issue and a democratic-governance issue. Those responsibilities frequently sit in different departments, different committees and even different levels of government, yet the technology will not respect any of those organizational boundaries.

A human resources department may see workforce disruption. A finance department may see changes in revenues and expenditures. An IT department may see cybersecurity and procurement. A school district may see educational opportunity and risk. A planning department may see data-center infrastructure, electricity demand or changes in its workforce. Each may understand its piece of the problem perfectly while nobody is responsible for understanding the whole. That is the governance problem we should be confronting now.

At the same time, government cannot view itself merely as the institution responsible for protecting people from artificial intelligence. It also has an obligation to help people benefit from it. Much of what citizens experience as government bureaucracy is ultimately an information problem. A resident is trying to understand a property-tax bill. A family is trying to determine whether it qualifies for assistance. A small business owner is navigating a permitting process. A public employee is searching thousands of pages of regulations. An elected official is trying to understand a hundred-page financial report. A small municipality is trying to analyze its pension liabilities or long-term capital needs without the analytical staff of a large city.

AI has the potential to dramatically lower the cost of expertise and make sophisticated analytical capabilities available to organizations and individuals that historically could not afford them. For local government in particular, that could be transformative. Thousands of public agencies across the country operate with very small staffs while facing increasingly complicated responsibilities in finance, infrastructure, pensions, cybersecurity, environmental compliance, grants and public communication.

If AI allows a city manager in a community of 8,000 people to access analytical capabilities that once required a much larger professional staff, that is more than an efficiency gain. It is an expansion of institutional capacity. Used well, AI could make government easier to navigate, faster to respond and better able to make decisions.

The harder question begins when we move from what AI can do to what AI means for the structure of the economy. If artificial intelligence eventually allows one person to accomplish what previously required several people, businesses will have powerful incentives to reorganize around that reality. The exact magnitude of the employment impact is impossible to know today, and technological history should make us cautious about predictions of mass unemployment. But we do not need certainty about the final number to recognize the economic pressure that is developing.

AI increasingly makes cognition scalable, and a large portion of the modern economy is built around paying people to perform cognitive work: accounting, legal analysis, customer service, programming, financial analysis, claims processing, underwriting, scheduling, research, documentation and countless other activities.

The consequences of disruption in those areas extend far beyond employment statistics because our fiscal system is deeply tied to human labor. Work is how most households earn income. Employment generates payroll taxes and income taxes. Those revenues finance governments, pensions, healthcare programs and social insurance. If capital and automated intelligence begin substituting for labor on a significant scale, governments could eventually face an uncomfortable imbalance: greater demand for retraining, income support and other transition assistance at precisely the same moment that parts of the traditional labor-based tax base are under pressure.

That is why I think AI should force a broader conversation about public finance. Our tax system was built for an economy in which human labor was central to economic production.

An employer hires workers, pays wages, pays payroll taxes, and those workers in turn pay income taxes and consume goods and services. But imagine an increasingly automated company that produces dramatically more output with dramatically fewer employees. Its productivity may increase substantially while its contribution through traditional labor-based taxation declines.

There is nothing inherently wrong with that company. It may simply be doing exactly what our economy encourages companies to do: becoming more productive. The problem is that the architecture of public finance may no longer align as neatly with the architecture of economic production. That does not mean we should rush to tax every AI model, robot or unit of computing power.

Poorly designed taxation could discourage productive innovation and become obsolete almost as quickly as it is enacted. But we should begin asking the underlying question now: if economic value increasingly comes from capital, computing and automated intelligence rather than human labor, should the way we finance public institutions gradually evolve as well?

That conversation will be difficult under any circumstances. It will be far more difficult if governments wait until they are simultaneously confronting weaker revenues and greater social costs.

There is also a more fundamental question that neither markets nor technology companies can answer for us. What should remain human? As AI becomes more capable, economic pressure will encourage us to automate whatever can be automated. But technical capability and social desirability are not the same thing. There are moments in public and private life when human presence, judgment and accountability have value independent of efficiency.

Teaching a child, caring for an elderly parent, telling a patient that she has cancer, exercising police authority, determining whether someone should lose access to a public benefit, or making a decision that significantly affects another person's liberty or livelihood are not merely information-processing tasks. AI may eventually become technically capable of performing substantial portions of all of them. That does not mean society must surrender those decisions to machines.

In many cases, the most appropriate model will probably be a partnership in which AI provides extraordinary analytical capability while a human remains responsible for judgment and accountability. The boundaries will change over time, but those boundaries should emerge from democratic choices rather than simply from procurement decisions or corporate optimization. Government exists not merely to maximize efficiency but to express collective judgments about what we value.

AI could either strengthen or weaken public trust. Local government provides a useful example because it is where citizens most directly experience government in their everyday lives. People encounter government through the road they drive, the park where their child plays, the emergency response when they call 911, the permit they need for their business, the water coming from their faucet and the taxes on their home.

If artificial intelligence helps public institutions answer questions more clearly, identify problems earlier, analyze finances more effectively and reduce unnecessary bureaucracy, citizens may experience government as more competent and responsive. At a moment when trust in institutions is already fragile, that matters.

But it is easy to imagine the opposite outcome. Residents unable to reach a human being, decisions generated by systems nobody can explain, employees deferring responsibility to algorithms, and citizens told that "the system" denied their request. That might produce administrative efficiency, but it would also be corrosive. Public institutions therefore cannot deploy AI exactly as a commercial organization might. Government has obligations of transparency, due process, and accountability. An algorithm cannot ultimately bear public responsibility. Someone must remain answerable for the decision.

The most difficult part of all of this will not necessarily be the world that exists once AI is fully integrated into the economy. It will be the transition from here to there. Technological optimists understandably focus on the extraordinary possibilities: cheaper healthcare, faster scientific discovery, individualized education, more productive businesses and dramatically more capable governments. I hope all of those things happen. But economies do not move instantaneously from one equilibrium to another. People live through the transition. A 23-year-old entering a labor market with fewer entry-level positions experiences technological change differently from an economist looking at aggregate productivity. So does a 52-year-old whose profession suddenly loses much of its market value, or a community whose largest employer discovers that automation allows it to operate with half the workforce.

The economy may ultimately become wealthier while millions of individual lives become more uncertain along the way. That is precisely the kind of collective challenge government exists to manage. Managing it does not mean trying to stop technological progress. Slowing AI development globally may not even be realistic given the economic and geopolitical incentives involved. It means recognizing that the benefits of technological progress and the costs of transition will not automatically fall on the same people, at the same time, or in the same places.

We need something broader than an AI policy. We need a public strategy for the AI transition.

Governments should already be asking how they will govern AI inside their own institutions, which public services could be dramatically improved by it, what workforce capabilities they will need in five or ten years, how exposed their revenues are to structural changes in employment, how they will respond if particular communities experience concentrated economic disruption, and which public decisions must retain meaningful human accountability.

They should also be asking how the benefits of AI will reach smaller communities, lower-income households and institutions that do not have the resources of major corporations or wealthy individuals. Perhaps most importantly, someone must be responsible for looking across all of these questions together. If every agency owns one piece of the problem, nobody owns the transition.

The technology industry will continue building increasingly capable artificial intelligence because the incentives to do so are enormous. Businesses will adopt it because competition will compel them to. Individuals will use it because it will make them more capable. Those forces are already operating. The unanswered question is what institutions we build around the productivity, disruption and new forms of power that AI creates.

That cannot be delegated to technology companies because it ultimately involves questions about taxation, work, education, public investment, human dignity, opportunity and accountability. Those are political questions in the broadest and best sense of the word.

Technology creates possibilities. Institutions determine how those possibilities are translated into society and how their benefits and burdens are distributed. Artificial intelligence may become one of the greatest expansions of human capability in history. Our challenge is not simply to invent it or regulate it. Our challenge is to build public institutions capable of using it well.

---

Andrew Flynn

Andrew Flynn writes about public leadership, fiscal stewardship, and the systems communities rely on to function well. He is a commissioner in Mt. Lebanon, Pennsylvania, works in public finance, and serves as a volunteer firefighter and EMT. Browse the Writing section for more articles, or visit Meet Andrew to learn more.