5 ways AI could strengthen democracy, not destroy it

Only 7% of young Americans view U.S. democracy as healthy, according to data cited in Beth Simone Noveck's new book "Reboot: AI and the Race to Save Democracy." Her argument: AI might be part of the fix, not just the problem. Noveck, who served as New Jersey's first Chief AI Strategist and advised the White House, 10 Downing Street, and the German Chancellery on technology policy, spent three years building AI tools with students and communities. The results, she says, point to AI applications that help governments find vulnerable people before crises hit.

What does 'democratic AI' actually look like?

Noveck's book opens with Sandricka Henderson, a 38-year-old single mother in Los Angeles with lupus. After the disease forced her to leave her job, her disability benefits barely covered rent. She was weeks from eviction.

Then a caseworker called her out of the blue. Henderson thought it was a scam. It wasn't.

Los Angeles built an AI system that analyzes emergency room visits, food assistance usage, and housing history to predict who is most likely to become homeless. Instead of waiting for people to arrive at shelters, the city contacts them before they lose their homes.

This is the distinction Noveck draws throughout: AI as a tool that helps a human caseworker find Sandricka in time, not a robot that replaces the caseworker. The system flags risk. A person makes the call.

The five applications Noveck documents

Beyond the LA homelessness prevention program, Noveck's book catalogs four other government AI deployments that are already operational:

• Massachusetts uses AI to identify federal benefit programs the state didn't know it qualified for, unlocking millions in unclaimed funds.
• Boston deployed AI to accelerate small business permit approvals.
• California launched an AI chatbot after the 2025 wildfires that provides wildfire resources in 70 languages.
• Multiple jurisdictions are using predictive models to allocate social services before crises escalate.

What unites these examples is a pattern: governments using AI to scan their own data and surface opportunities or risks that human staff would miss, then routing those findings to human workers who act on them.

Why this argument matters right now

The dominant narrative around AI and democracy runs the opposite direction. Headlines warn about AI-generated disinformation, deepfakes of candidates, and the collapse of shared truth. Noveck acknowledges these risks but argues the coverage has crowded out a different story: AI as boring infrastructure that helps bureaucracies work.

Her credentials matter here. Noveck founded and directed The GovLab at NYU, one of the most cited institutions in civic technology research. She's not a tech evangelist parachuting into policy debates. She's spent two decades in the gap between what governments promise and what they actually deliver.

“Recent headlines warn of brain fry from AI use and an AI apocalypse on par with nuclear war. It could seem we are living in the worst of times.”

— Beth Simone Noveck, Professor at Northeastern University

Her counterpoint is that the apocalyptic framing misses where AI is already changing government operations. The Los Angeles homelessness system isn't speculative. Massachusetts is already collecting federal money it didn't know existed. Boston's permits are already faster.

These aren't glamorous applications. They don't make headlines about superintelligence or existential risk. But they address something measurable: whether a single mother keeps her housing, whether a small business opens in weeks instead of months.

The missing piece: what the book doesn't prove

Noveck's examples are persuasive but narrow. Each case shows AI helping government do something it already wanted to do, faster or more precisely. The harder question is whether these tools can address the structural problems driving the 7% figure: polarization, distrust, the sense that institutions don't represent people's interests.

A faster permit process doesn't heal the wound when people believe elections are rigged or that their vote doesn't matter. A multilingual chatbot is useful but doesn't rebuild civic trust.

Noveck's book, based on the excerpts shared, appears to argue that competent government creates trust, and AI can make government more competent. That's a defensible theory. But the chain from "my housing case was handled well" to "I believe in democracy" has links the book may not fully establish.

What this means for AI teams building public sector tools

The pattern in Noveck's examples suggests a narrow but achievable product strategy: AI that helps government find people who need services, surfaces opportunities government is missing, or accelerates processes that are slow because humans can't scan enough data.

This is different from AI that makes decisions or replaces workers. The political economy of those tools is brutal. Unions resist, procurement officers hesitate, and any failure becomes a headline. But AI that makes a caseworker's existing job more effective? That sells.

The California wildfire chatbot example is instructive. Launching in 70 languages would be impossible with human translators on a wildfire's timeline. The AI solved a specific scaling problem that no one could object to.

Whether these tools add up to "saving democracy" is a separate question. But they add up to a product category with real traction and measurable ROI, which is where most govtech conversations actually begin.

Source: Fast Company / Next Big Idea Club