Before Public Institutions Bet on AI, They Need a Public Value Test
Artificial intelligence is moving quickly into public life. Governments, nonprofits, schools, and employers are being told that AI can make systems faster, cheaper, and more efficient. Some of that may be true. But efficiency cannot be the only measure that matters, especially when public institutions are making decisions that affect workers, communities, public services, and democratic accountability.
That is why I believe public institutions need a Public Value Test for AI investment.
Right now, many conversations about AI adoption begin with the technology itself: what it can automate, how much time it can save, or how quickly it can be deployed. But public institutions have a broader responsibility. They have to ask what these tools will mean for the people who use public services, the workers who deliver them, and the communities that live with the long-term consequences of these decisions.
This is especially important because AI investment is not neutral. Public procurement, tax incentives, infrastructure approvals, workforce funding, and economic development strategies all help shape where capital flows. If those decisions are made without clear public standards, AI investment can deepen dependence on a small number of firms, redirect resources away from other urgent needs, and weaken accountability inside public systems.
A Public Value Test would help governments and civic institutions ask better questions before AI becomes embedded in their work.
Does the proposed AI investment improve the quality of public services, or does it mainly reduce labor costs? Does it support workers and improve job quality, or does it increase surveillance, displacement, and work intensity? Does it reduce inequities, or does it reproduce them through opaque systems and unequal access? Can affected people understand and challenge how the system is being used? Does the investment build long-term public capacity, or does it make agencies more dependent on private vendors?
These questions should be asked before public money, public authority, or public infrastructure is used to support AI adoption.
A Public Value Test: Five Areas
First, service quality. AI tools should be evaluated based on whether they improve real outcomes for people, not only whether they move cases faster or cut administrative costs.
Second, labor impact. Public agencies and employers should assess how AI affects job quality, bargaining power, skill development, workload, and employment stability.
Third, equity. AI investments should be examined for their effects across race, class, gender, geography, disability, immigration status, and other lines of inequality.
Fourth, transparency and contestability. People affected by AI systems should know when these tools are being used and have meaningful ways to challenge mistakes, harms, or unfair decisions.
Fifth, institutional capacity. Governments should ask whether AI adoption strengthens public expertise and democratic control, or whether it locks agencies into long-term dependency on private firms.
This is not an argument against innovation. It is an argument for making innovation accountable to public goals.
AI will shape labor markets, public budgets, infrastructure, and the future of service delivery. The question is whether those changes will concentrate wealth and power or help build a more democratic and broadly shared economy. Public institutions have tools they can use: procurement standards, labor-impact assessments, subsidy conditions, community benefit requirements, public-interest research funding, antimonopoly enforcement, and auditability rules.
But those tools matter most when they are used early.
Before governments bet on AI, they should require AI investments to demonstrate public value. Speed is not enough. Innovation is not enough. The real test is whether these systems help people live with more security, dignity, agency, and democratic control over the institutions that shape their lives.