The Reluctant Modernizer
Government is where innovation goes to slow down.
This isn't entirely a flaw. When the Department of Motor Vehicles takes three weeks to process your paperwork, it's frustrating. When the FDA takes three years to approve a drug, lives hang in the balance. But government slowness also reflects something important: the stakes of getting things wrong when the government acts are different from when a startup fails. Government decisions affect everyone, often involuntarily. Caution has value.
Yet the gap between what government can do and what the private sector can do has widened dramatically. You can transfer money internationally in seconds, but changing your address with the government takes days. You can get a mortgage approval in minutes, but a building permit takes months. You can order anything and have it delivered tomorrow, but renewing a passport takes weeks.
The century from 1926 to 2026 saw the expansion of government to an unprecedented scale and scope—and the corresponding growth of bureaucracy, regulation, and administrative process. The same period saw digitization transform every other sector while government lagged stubbornly behind.
This chapter traces that history: the growth of the administrative state, the uneven digitization of government services, the rise of surveillance and cybersecurity challenges, and the tension between efficiency and accountability that makes government transformation so difficult. Understanding this context is essential for understanding what AI might change—and what forces might resist that change.
2026 Snapshot — Government Today
Scale and Scope
Government spending: Developed countries: 35-55% of GDP. US federal spending ~24% of GDP; state and local add another ~15%.
Employment: US federal civilian employees ~2.9 million; state and local ~20 million. Total government employment ~15% of workforce in developed countries.¹
Regulation: US Federal Register ~90,000 pages annually. Code of Federal Regulations exceeds 185,000 pages. EU acquis communautaire over 170,000 pages.
Services: Social security, healthcare (Medicare, Medicaid, NHS), education, infrastructure, defense, justice, permits, licensing, records.
Digitization Status
Highly digitized:
- Tax filing (IRS e-file: 90%+ of returns)
- Social security benefits (online enrollment, direct deposit)
- Some permits and licenses (varies by jurisdiction)
- Voter registration (many states)
Partially digitized:
- Courts (electronic filing growing; paper persists)
- Healthcare administration (EHR mandates; interoperability limited)
- Immigration (online applications; processing still slow)
Barely digitized:
- Many local permits (building, zoning—often paper)
- Criminal justice records (fragmented, inconsistent)
- Inter-agency information sharing (siloed systems)
Trust and Legitimacy
Declining trust: Public trust in government at historic lows in many democracies. US trust in federal government ~20% (vs. 75% in 1960s).²
Legitimacy challenges: Polarization; misinformation; election integrity concerns; perceived corruption; responsiveness failures.
Delivery gaps: Long wait times, inconsistent service, hard-to-navigate processes feed distrust.
Technology Challenges
Legacy systems: Critical government systems run on COBOL, mainframes from the 1960s-80s. IRS, Social Security, state unemployment systems.
Cybersecurity: Government targets for nation-state hackers. OPM breach (2015) exposed 21 million records. Ongoing attacks on critical infrastructure.
Procurement: Slow, complex processes favor large incumbents. Average federal IT project takes 4+ years; failure rate high.
Interoperability: Systems don't talk to each other. Same information requested repeatedly.
Notable Players
Government Technology Leaders
Estonia: E-Residency, digital identity, X-Road data exchange. 99% of government services online. Often cited as digital government model.³
Singapore: GovTech agency; national digital identity (SingPass); integrated services. High trust, small scale, different political context.
Denmark, Finland, Norway: Consistently ranked high in digital government indices. Strong digital identity infrastructure.
United Kingdom: Government Digital Service (GDS) established 2011; GOV.UK consolidated 1,900+ websites; service design standards.
United States: US Digital Service (2014); 18F; Login.gov digital identity. Progress but fragmentation across agencies and states.
GovTech Companies
Palantir: Data integration and analysis for government. Controversial; used by intelligence, law enforcement, immigration.
Tyler Technologies: Local government software—courts, public safety, civic services. Largest US public sector software company.
Salesforce Government Cloud, Microsoft Government, AWS GovCloud: Cloud platforms adapted for government compliance requirements.
Civic tech startups: Code for America; Nava PBC; Civic Actions; smaller govtech firms addressing specific problems.
International Organizations
OECD: Digital Government Index; policy frameworks.
World Bank: ID4D (Identification for Development); digital government capacity building.
UN: E-Government Survey; sustainable development goals.
The Century in Government
The Rise of the Administrative State (1920s–1970s)
The New Deal expansion:
- Social Security (1935)
- Securities regulation (SEC 1934)
- Labor standards (NLRA 1935, FLSA 1938)
- Banking regulation (FDIC, Glass-Steagall)
Government went from ~3% of US GDP (1929) to 20%+ post-WWII.
Post-war consolidation:
- Medicare and Medicaid (1965)
- Environmental protection (EPA 1970, Clean Air Act, Clean Water Act)
- Occupational safety (OSHA 1971)
- Consumer protection expansion
The regulatory state emerges: Independent agencies; notice-and-comment rulemaking; administrative procedure becomes field of law.
Digitization Begins (1970s–2000s)
Early computing:
- Social Security Administration among first large-scale government computerizations (1950s-60s)
- IRS computer systems (1960s)
- Military and intelligence systems
The PC era:
- Desktop computers reach government offices (1980s)
- Databases replace paper files (partially)
- Email begins to supplement memos
The internet arrives:
- FirstGov.gov (2000) as government portal
- E-file for taxes (1990s)
- Online information access expands
- Government websites proliferate (often poorly)
The Security State (2001–Present)
Post-9/11 expansion:
- Department of Homeland Security created (2002)
- USA PATRIOT Act expanded surveillance authorities
- TSA, enhanced border security
- Intelligence community expansion
Mass surveillance revealed:
- Snowden disclosures (2013) showed scope of NSA programs
- Ongoing tensions between security and privacy
- Encryption debates
Cybersecurity becomes critical:
- OPM breach (2015): 21.5 million records
- SolarWinds attack (2020): widespread government compromise
- Ransomware attacks on state and local governments
- Critical infrastructure protection concerns
Digital Government Push (2010s–Present)
Institutional innovation:
- UK GDS (2011) and GOV.UK
- US Digital Service (2014) and 18F
- Similar units in many countries
Design thinking enters government:
- User-centered service design
- Agile development methods
- API-first approaches
- Cloud migration begins
Results mixed:
- Some successes (UK digital transformation, Healthcare.gov recovery)
- Many failures (state unemployment systems during COVID)
- Legacy systems persist; modernization slow
Modern Bottlenecks
Legacy Technology
The COBOL problem: Critical systems run on 60-year-old technology. Social Security, IRS, state unemployment systems. Few programmers know COBOL; systems are fragile and hard to modify.
Technical debt: Decades of patches and workarounds. Systems that no one fully understands. Documentation lost.
Procurement barriers: Multi-year contracts; requirements written before technology decided; incumbent advantages; risk aversion.
Integration failures: Systems don't interoperate. Data exists but can't be shared. Same information requested repeatedly.
Institutional Constraints
Civil service rules: Hiring processes take months. Pay scales don't compete with private sector for technical talent. Firing is difficult.
Budget cycles: Annual appropriations prevent long-term planning. "Use it or lose it" spending. Capital vs. operating distinctions.
Political cycles: Administration changes reset priorities. New initiatives abandoned; ongoing projects reprioritized.
Risk aversion: Failure is punished more than success is rewarded. Innovation is risky; maintaining status quo is safe.
Process Requirements
Administrative procedure: Rulemaking takes years. Notice, comment, response, litigation. Designed for accountability; creates delay.
Compliance burdens: Privacy, security, accessibility, procurement rules all add requirements. Each reasonable; cumulative effect is paralysis.
Documentation and accountability: Everything must be documented. Audit trails, records retention, FOIA. Paper persists because it's defensible.
Trust Deficit
Why does this matter?: Low trust means government initiatives face skepticism. Modernization seen as threat, not benefit. Privacy concerns amplified.
Self-reinforcing: Poor service reduces trust; low trust makes modernization harder; slow modernization perpetuates poor service.
Courts and Justice
Current State
Backlogs: US federal courts: 760,000+ pending cases. State courts: millions more. Cases take years to resolve.
Inequality: Quality of justice depends on resources. Public defenders overwhelmed; private attorneys expensive. Bail systems perpetuate inequality.
Inconsistency: Different judges, different outcomes. Sentencing varies. Geography matters.
Technology lag: Courts were among the last to adopt electronic filing. Video hearings became widespread only with COVID.
Recent Changes
Electronic filing: Now widespread in federal courts and many state courts. Accelerated by pandemic.
Virtual hearings: Routine for many proceedings now. Access improved; due process concerns remain.
Case management systems: Digital dockets, scheduling, tracking. Implementation varies.
Legal research: AI-assisted research (Westlaw, LexisNexis with AI features). Lawyers use but courts slower to adopt.
Ongoing Challenges
Evidence and discovery: Digital evidence creates new challenges. E-discovery is expensive and complex. Chain of custody for digital information.
Bias in algorithms: Pretrial risk assessment tools face criticism. Sentencing algorithms controversial. "Garbage in, garbage out" plus opacity.
Access to justice: Most civil legal needs go unmet. Self-represented litigants struggle. AI could help or create new barriers.
Identity, Voting, and Civic Infrastructure
Identity Systems
Current state: Fragmented. US has no national ID; Social Security numbers used as de facto identifier despite not being designed for it. Driver's licenses vary by state.
Problems: Identity theft; fraud; difficulty proving identity; barriers for those without documents.
Digital identity: SSA; Login.gov (federal); state DMV; private sector (banks, credit bureaus). Not interoperable.
Other countries: Estonia X-Road and digital ID; India Aadhaar (1.3 billion enrolled); EU eIDAS framework.
Voting Systems
Current state: Decentralized (US has 10,000+ election jurisdictions). Mix of technologies: paper, optical scan, DRE machines.
Security concerns: Nation-state targeting; disinformation; confidence undermined regardless of actual security.
Accessibility vs. security: Online voting convenient but security risks high. Paper ballots more secure but less accessible.
Audit trails: Paper records increasingly required. Risk-limiting audits gaining adoption.
Information Integrity
The challenge: Synthetic media makes truth harder to verify. Deepfakes, generated content, manipulated images.
Government role: Provenance and authenticity infrastructure. Content authentication. Public records verification.
Current tools: C2PA (Coalition for Content Provenance and Authenticity); content credentials; digital signatures.
The AI Transformation Beginning
Current Government AI Use
Fraud detection: IRS, Social Security, unemployment programs use pattern recognition. Mixed effectiveness; false positives problematic.
Document processing: Immigration, patent office, courts using AI for document analysis. OCR and classification.
Customer service: Chatbots for routine inquiries. Generally basic; human escalation common.
National security: Intelligence analysis; surveillance; defense applications (classified).
Predictive applications: Pretrial risk assessment (controversial); benefits eligibility; audit selection.
Emerging Applications
Administrative automation: Routine decisions that currently require human review. Permit approvals, benefits determinations, compliance checking.
Citizen services: AI assistants for navigating government. Translation, summarization, guidance.
Policy analysis: Simulation and modeling for policy design. Impact prediction.
Compliance: Real-time compliance monitoring. Machine-readable regulation.
Concerns and Constraints
Accountability: When AI makes decisions, who is responsible? Administrative law assumes human decision-makers.
Bias and fairness: AI systems can perpetuate or amplify existing biases. Disparate impact on protected groups.
Transparency: Black-box decisions conflict with due process. Citizens have right to understand how decisions are made.
Security: AI systems create new attack surfaces. Adversarial attacks, data poisoning, model compromise.
The Path Forward
Near-Term Likely (2026-2032)
Digital services expand: More routine government interactions available online. Mobile-first design. Improved user experience.
AI augments human workers: Chatbots handle routine inquiries; AI assists case workers; document processing automated.
Legacy systems face crisis: COBOL programmers retire. Systems increasingly fragile. Some failures force emergency modernization.
Cybersecurity investment increases: After more breaches and attacks. Zero trust adoption. Critical infrastructure protection.
Courts adopt more technology: Electronic filing universal; AI for research and scheduling; virtual hearings routine for appropriate cases.
Plausible (2032-2040)
Machine-readable regulation: Some rules encoded for automated compliance checking. Real-time permit processing for routine cases.
Digital identity matures: Interoperable digital ID accepted across jurisdictions. Privacy-preserving credentials enable selective disclosure.
AI case processing: Routine administrative decisions largely automated with human oversight. Faster, more consistent.
Courts transform: AI assists judges with research, drafting, scheduling. Case prediction helps settlement. Online dispute resolution expands.
Government as platform: APIs enable private sector to build services on government data and functions.
Wild Trajectory (2040+)
Algorithmic government: Most routine government functions automated. Human role shifts to oversight, policy, and exceptions.
Real-time compliance: Regulation is code. Compliance is verified continuously. Violations detected automatically.
Radical efficiency: Government services delivered in minutes, not months. Friction approaches zero for routine interactions.
New governance questions: What is the role of human judgment? How are algorithms accountable? Who audits the auditors?
Risks and Guardrails
Automation Bias
Risk: Humans defer to AI recommendations without adequate scrutiny. "The computer says" becomes excuse for bad decisions.
Guardrails: Meaningful human review; training on AI limitations; accountability for human decision-makers regardless of AI recommendation.
Algorithmic Discrimination
Risk: AI systems encode and amplify existing biases. Protected groups systematically disadvantaged.
Guardrails: Bias audits before deployment; ongoing monitoring; disparate impact testing; transparency about how systems work; appeal mechanisms.
Loss of Human Touch
Risk: Efficiency gains come at cost of responsiveness to individual circumstances. Edge cases badly served.
Guardrails: Human escalation paths; discretion preserved for complex cases; design for exceptions not just average cases.
Security and Attack Surface
Risk: AI systems create new vulnerabilities. Adversarial inputs, data poisoning, model compromise.
Guardrails: Security by design; red-teaming; input validation; monitoring for anomalous behavior; defense in depth.
Democratic Accountability
Risk: Decisions made by AI are hard to contest, understand, or change through democratic processes.
Guardrails: Transparency requirements; explainability; legislative oversight of AI deployment; public participation in AI governance.
Surveillance Expansion
Risk: AI enables surveillance at scale that wasn't previously possible. Privacy erodes; chilling effects on behavior.
Guardrails: Purpose limitations; data minimization; oversight mechanisms; legal constraints on government surveillance.
The Deeper Questions
Efficiency vs. Accountability
Government is slow partly because it's accountable. Every shortcut creates risk of abuse, error, or unfairness. Every process exists because someone, at some point, was harmed by its absence.
AI offers efficiency—but can it preserve accountability? Can algorithmic decisions be meaningfully reviewed? Can affected citizens understand and contest them?
The answer isn't obvious. But the question is essential.
What Should Government Do?
As AI transforms private sector, it will transform what government can do. But should government do everything it can?
Surveillance that was impossible becomes easy. Prediction that was speculative becomes reliable. Decisions that required human judgment can be automated.
The capability question is separate from the should question. AI enables a more capable state—for better and worse.
Trust as Infrastructure
Government works when citizens trust it. Trust is built through experience: services that work, decisions that are fair, responsiveness to problems.
AI could build or destroy trust. Efficient services might restore faith. Opaque decisions might further erode it. The implementation matters more than the technology.
Conclusion
Government has grown enormously over the past century—in spending, employment, regulation, and reach. The administrative state that emerged from the New Deal and expanded through subsequent decades now touches nearly every aspect of life.
Yet government has been slow to adopt the technologies that transformed every other sector. Legacy systems, institutional constraints, process requirements, and risk aversion combine to create a widening gap between what government can do and what citizens experience elsewhere.
AI offers the possibility of transformation—faster services, more consistent decisions, better information. But AI also creates risks: automation bias, algorithmic discrimination, loss of human judgment, security vulnerabilities, and democratic accountability challenges.
The chapters that follow explore specific transformations: regulation as code, courts and justice, identity and voting, and the bureaucratic functions that might be streamlined. Each involves the same tension: efficiency gains against accountability risks, capability expansion against privacy concerns, modernization needs against institutional resistance.
Government will be among the last institutions to transform. It will also be among the most consequential. How society navigates this transformation will determine whether AI makes government more responsive to citizens or more removed from them.
Endnotes — Chapter 32
- US federal civilian employment approximately 2.9 million (OPM data); state and local government employment approximately 20 million (BLS data).
- Pew Research Center public trust in government data shows decline from ~75% in 1960s to ~20% in recent years, with variation by administration and events.
- Estonia digital government includes X-Road data exchange, digital identity, e-Residency program; 99% of government services available online.
- US Federal Register pages vary annually; Code of Federal Regulations total pages exceeds 185,000 as of recent counts.
- OPM breach (2015) compromised 21.5 million records including security clearance information; attributed to Chinese state-sponsored hackers.
- Healthcare.gov initial failure (2013) and recovery demonstrates both failure modes and potential of digital service teams.
- COBOL systems still process 95% of ATM transactions and 80% of in-person transactions; many government systems run on COBOL with retiring workforce.
- UK Government Digital Service established 2011; GOV.UK consolidated approximately 1,900 government websites.
- India Aadhaar biometric identity program has enrolled over 1.3 billion people; largest biometric ID system globally.
- Pretrial risk assessment tools (COMPAS, PSA, Arnold) controversial; ProPublica investigation (2016) raised concerns about racial bias in COMPAS predictions.