Public administration & e-government

In-depth analysis

KI1 — One-stop digital administrative services

• Mechanism: Creation of a unified Government Gateway API layer that integrates the existing back-end systems (EESZT — the national e-health infrastructure, NAV Online — the tax authority’s online platform, Ügyfélkapu+ — the upgraded Hungarian “Client Gate” e-government portal) behind a single front end. With a single login (eIDAS-compatible electronic identification), the citizen reaches every service. The “once-only” principle applies in the background: whatever the state already knows about a citizen is not requested again — institutions share data via APIs.
• Quantified target: By 2028, 100% of the 25 most frequent case types are available online; the average processing time drops from 30 days to 5 working days; in-person traffic at government windows (kormányablak) falls by 60%.
• International precedent: Estonia’s X-Road system (since 2001): unified data-exchange infrastructure among 900+ organisations, with 99% of citizens handling their cases online. Denmark (NemID/MitID): in 2023 the digital case-handling rate among the adult population was 92%.
• Trade-off / risk: The “single point of failure” problem: the outage of a single platform paralyses all administrative services (cf. the 2021 Estonian digital identification crisis, when an ID-card chip flaw affected 750,000 citizens). Moreover, the older, digitally less literate population (65+: ~2M people) may be excluded if there is no parallel offline channel.

KI2 — AI assistant for administrative services

• Mechanism: A natural-language-processing (NLP)-based chatbot which, linked to KRÉTA-type rule engines, not only informs but also pre-fills forms, validates data before submission, and sends proactive notifications (e.g. “your driving licence has expired, you can renew it in 2 clicks”). It applies choice-architecture principles: smart default values (e.g. the most frequently chosen option pre-selected), simplified choice sets, warnings at typical error points.
• Quantified target: The share of rejected/incomplete submissions falls from 35% to below 10%; in 70% of cases the AI assistant can handle the entire process without human intervention; customer satisfaction (CSAT) > 4.0/5.0.
• International precedent: Singapore’s “Ask Jamie” virtual assistant covers 70+ government agencies with 1M+ interactions per month. Portugal’s ePortugal chatbot cut telephone customer-service calls by 40% in 2022.
• Trade-off / risk: The “hallucination” problem: if the AI gives erroneous legal guidance (e.g. suggests a wrong deadline or the wrong form), the citizen may suffer a loss of rights. Rule-engine validation and a “confidence threshold” are required — if the AI is uncertain, it routes to a human case handler. Data-protection risk is also significant: the AI needs access to the citizen’s personal data.

KI3 — Bureaucracy reduction — measurably

• Mechanism: Application of the Standard Cost Model (SCM): measuring the cost of every administrative obligation (time × hourly wage × number of those affected × frequency). An annual “Bureaucracy Audit” is mandatory for every ministry: how many forms were eliminated, how many data requests were replaced by automatic data exchange, how much cumulative time was saved. Results are published on a public dashboard with a per-ministry ranking.
• Quantified target: Over 4 years, total administrative burden falls by 25% (measured in SCM); the number of active government forms falls by 30%; the “Simple State” index (a proprietary metric: average number of administrative steps × average number of days) improves by 10% annually.
• International precedent: The Netherlands’ “ACTAL” system (2000–2017): the SCM methodology achieved a 25% administrative-burden reduction over 8 years. South Korea’s “Government 3.0” programme: customer satisfaction rose by 20 percentage points after the public measurement system was introduced.
• Trade-off / risk: The metric-optimisation hazard (Goodhart’s law): if ministries cut the number of forms but make the remaining ones more complex, the actual burden may grow. “Citizen journey”-based measurement is also needed (how long it takes a citizen to go through an administrative case from start to finish), not just input counting.

KI4 — Municipal digitalisation

• Mechanism: A central “SaaS government” model: the state provides the software platform (cloud-based, modular) onto which municipalities connect. They do not need to maintain their own IT teams. The system runs on a unified base (e.g. building permit, social application, local tax) but also allows local customisation. Roll-out is supported by “digital mentors”: trained experts who provide on-site assistance.
• Quantified target: By 2027 at least 80% of the 3,200 municipalities connect to the unified platform; the share of local online case handling rises from 15% to 50%; the “digital mentor” programme reaches 500 settlements in the first 2 years.
• International precedent: Finland’s “Suomi.fi” platform: a central digital-services layer onto which municipalities connect — the online case-handling rate rose to 78%. Australia’s “GovCMS”: a shared content-management platform for 200+ government bodies, cutting IT costs by 40%.
• Trade-off / risk: Loss of local specificity: if the central system is too rigid, municipalities cannot handle specific local cases (e.g. a mining town and a suburban commuter city generate very different needs). Vendor lock-in risk: if the system depends on a single supplier, that supplier can dictate prices. An open-source platform is to be preferred.

KI5 — Behavioural public policy unit (Nudge Unit)

• Mechanism: A 15–20-person interdisciplinary team (behavioural economists, psychologists, data analysts, UX designers) under the Prime Minister’s Office or the ministry for public administration. Three main workstreams: (1) “Regulatory Review”: mandatory behavioural-science review of every new regulation and form, (2) “RCT Pipeline”: 10–15 randomised controlled trials (A/B tests) per year on citizen-facing interfaces, (3) “Nudge Report”: an annual public report on intervention impacts, with effect size and cost-benefit analysis.
• Quantified target: 10 RCTs run in the first year; within 3 years at least 5 interventions with a verified effect size of >5 percentage points improvement in the respective domain; per the “Nudge Report”, the average ROI of interventions is > 10:1 (every HUF 1 invested yields at least HUF 10 in savings).
• International precedent: UK Behavioural Insights Team (2010–): rewording tax-reminder letters raised tax revenues by GBP 5 billion over 10 years. A social-norm message (“9 out of 10 of your neighbours have already paid their tax”) increased on-time payment by 15%. The introduction of opt-out organ donation in Wales (2015) raised registrations by 25%.
• Trade-off / risk: Ethical concern: nudging borders on manipulation, especially when the state applies it to its own citizens. One must remain within the frame of “libertarian paternalism”: freedom of choice is preserved, opt-out is always available. Also, the “publication bias” risk: if only successful experiments are published, the Nudge Unit’s effectiveness may be overstated. Every experiment — successful or not — must be published publicly.

KI6 — Competitive public-service pay system

• Mechanism: On the Singapore model: civil-service and ministerial pay are pegged to the median private-sector pay for key positions (e.g. a 70% ratio). Lee Kuan Yew: “Underpaid ministers and public officials have ruined many governments in Asia. Adequate remuneration is vital for high standards of probity.” Underpaid civil servants are more prone to corruption, while talented people leave public administration.
• Quantified target: Civil-service pay rises to 70% of the private-sector median within 5 years; civil-service turnover falls by 20%; the number of corruption cases in public administration falls by 30%. 📖 Source: Lee Kuan Yew: From Third World to First (Chapter 12)

KI7 — Official selection and rotation system

• Mechanism: Kautilya’s Arthashastra describes the system of appointing and monitoring officials in detail: mandatory reference checks, a “character audit” before appointment, regular rotation to prevent the concentration of power, and an obligation for the official to prove that their wealth is proportionate to their income. Modern adaptation: five-yearly position rotation in corruption-prone areas (public procurement, permitting, regulatory inspection).
• Quantified target: The rotation system covers 100% of corruption-prone positions by 2030; annual wealth-proportionality review; the detection rate of suspicious wealth accumulation doubles. 📖 Source: Kautilya: Arthashastra (Book II)

KI8 — Drucker-based effectiveness measurement in public administration

• Mechanism: According to Drucker, effectiveness can be learned. A public-administration KPI system based on Drucker’s five practices: (1) customer-oriented goal setting, (2) results-based evaluation instead of time-spent evaluation, (3) contribution-driven mindset, (4) decision journal (what I decided, why, and what the outcome was), (5) strengths-based task allocation.
• Quantified target: KPI system introduced in every ministry by 2028; annual measurement of the customer satisfaction index; the decision journal documents 100+ key decisions per year. 📖 Source: Drucker: The Effective Executive (Introduction)

KI9 — Local participatory budgeting

• Mechanism: Tocqueville called local self-government the school of democracy: “Municipal institutions constitute the strength of free nations.” Citizens directly decide on part of the municipal budget (5–10%) in an annual participatory budgeting process. Both online and offline participation are ensured.
• Quantified target: By 2030, 100+ settlements with participatory budgeting; the affected budgetary envelope is HUF 20+ billion per year; the participation rate in affected settlements is 15%+. 📖 Source: Tocqueville: Democracy in America (Volume I, Chapter 5)

KI10 — “Once-only Plus” — proactive state service

• Mechanism: The state proactively contacts the citizen when it detects an entitlement (e.g. social benefit, pension supplement, tax allowance). Kautilya: “In the happiness of his subjects lies his happiness, in their welfare, his welfare.” It should not be for the citizen to know what they are entitled to — the state offers the service.
• Quantified target: By 2030, 60% of state benefits are proactively offered; application burden falls by 40%; entitlement coverage (those actually claiming / those entitled) rises from 70% to 90%. 📖 Source: Kautilya: Arthashastra (Introduction)

KI11 — Organisational behaviour audit — Allison framework

• Mechanism: The central insight of Model II (Organizational Behavior) in Allison and Zelikow’s Essence of Decision: organisations do not execute the political leadership’s intentions but follow their own routines and SOPs (Standard Operating Procedures) — organisational logic lives a life of its own. Modern public-administration application: an annual “Organisational Behaviour Audit” at key institutions (NAV — the tax authority, NEAK — the health insurance fund, NFSZ — the national employment service, GVH — the competition authority): (1) identify where organisational routines distort political intent, (2) where organisational culture impedes innovation, (3) where bureaucratic bargaining (Model III) leads to distortions in decision-making.
• Quantified target: By 2028, an Organisational Behaviour Audit at 5+ institutions per year; corrective action plan for 50% of the identified organisational distortions; audit results are public.
• Trade-off / risk: The audit is politically sensitive — institutional heads defend the status quo. The solution: the audit is conducted by an external, independent team, and the results are public. 📖 Source: Allison & Zelikow: Essence of Decision (Model II: Organizational Behavior)

KI12 — Cognitive-load-reducing administrative services

• Mechanism: The World Bank’s WDR 2015 demonstrated: poverty imposes a cognitive burden, and when designing citizen services one must take into account that the cognitive capacity of the neediest clients is limited. Programme: (1) every social and administrative form is no more than 1 page, with visual decision trees, (2) “smart defaults”: the AI assistant (KI2) pre-fills the most likely answer based on the client’s data, (3) “cognitive load audit”: before any new administrative process is introduced, it is tested with a disadvantaged target group — if the process is too complex, it must be simplified.
• Quantified target: By 2029, every social-benefit application form is no more than 1 page; 10+ cognitive load audits per year; the share of rejected/incomplete applications falls by 50% in disadvantaged target groups.
• Trade-off / risk: Simplification may increase the burden on the administrative side (more back-end processing). The “smart default” risk: if miscalibrated, the client may submit the application with erroneous data. 📖 Source: World Bank: World Development Report 2015 — Mind, Society, and Behavior (Chapters 4, 6)