Automated tax systems often raise compliance and cut administration costs, but their benefits are conditional: low digital literacy, weak trust and unclear regulation erode gains and can trigger taxpayer resistance.

Main Finding

Automation (including e‑filing, e‑invoicing, AI risk‑profiling, real‑time reporting and RPA) has generally improved tax administration efficiency and increased measurable compliance — but its behavioural effects are heterogeneous and heavily mediated by digital literacy, institutional trust, regulatory clarity and system usability. Automation can both enable voluntary compliance through simplification and fairness, and strengthen enforced compliance via data‑driven monitoring, yet risks widening digital divides and provoking resistance where governance, transparency or support are weak.

Key Points

• Evidence base: systematic review of 36 peer‑reviewed articles (2015–2025) across diverse country contexts.
• Dual role of automation:
  • Enabler of voluntary compliance: simplified processes (pre‑filled returns, e‑filing, integrated digital ID) reduce cognitive/administrative costs and increase taxpayer satisfaction and voluntary reporting (e.g., Peru’s e‑invoicing → >5% increase in VAT declarations within one year).
  • Enforcer of compliance: AI and ML risk profiling, real‑time data matching and automated audits raise perceived detection probability and enable targeted, low‑cost nudges (notifications, reminders).
• Behavioural heterogeneity:
  • Younger/digitally fluent taxpayers respond positively; older/low‑literacy taxpayers may experience anxiety or alienation.
  • Some taxpayers comply out of civic duty when systems are perceived as fair; others comply defensively when systems are perceived as opaque or punitive.
• Mediators/moderators:
  • Positive mediators: digital literacy, IT capacity within firms, usability, clear regulation, data protection, transparent processes.
  • Negative moderators: weak institutional capacity, poor infrastructure, opaque automated decision‑making, perceived inequity in enforcement.
• Illustrative national findings: Estonia (high integration with digital ID → higher satisfaction/compliance), Brazil (algorithmic nudges ↑ timely submission), Morocco (blockchain VAT trail reduced invoice fraud), Indonesia (implementation struggles due to infrastructure and transparency gaps).
• Theoretical framing: synthesis of deterrence theory (increased detection probability, ambient enforcement) and institutional theory (procedural fairness, legitimacy, trust).

Data & Methods

• Method: systematic literature review following PRISMA 2020 guidelines.
• Timeframe & scope: peer‑reviewed, English‑language journal articles published 2015–2025.
• Databases searched: Scopus, Web of Science, ProQuest.
• Search query (keywords combined): automation/robotic process automation/digital transformation/AI/machine learning/etc. AND tax compliance/taxpayer behaviour/tax morale/etc. AND influence/effect/impact/relationship/association.
• Screening numbers:
  • Initial hits: 215
  • Duplicates removed: 48 → 167 screened
  • Title/abstract exclusions: 83
  • Full texts assessed: 84
  • Full‑text exclusions: 41 (not directly focused on tax compliance n=13; not in English n=5; outside timeframe n=23)
  • Final included studies: 36
• Analysis: iterative thematic synthesis; articles categorized into four themes — (1) automation & voluntary compliance, (2) automation & enforced compliance, (3) behavioural responses to automated systems, (4) mediators & moderators.
• Limitations noted by the paper: reliance on published studies (heterogeneous methods and contexts), limited causal/longitudinal microdata across many settings, and potential publication bias toward positive automation outcomes.

Implications for AI Economics

• Revenue and compliance dynamics:
  • Automation can expand the effective tax base and increase reported liabilities quickly (e.g., e‑invoicing), altering short‑run revenue projections and improving tax collection efficiency.
  • Real‑time reporting and algorithmic audits change the timing and predictability of revenue flows and may reduce reliance on intensive audit staffing.
• Cost–benefit and allocation:
  • AI and automation can substitute for costly manual enforcement, but upfront investments in digital infrastructure, algorithm development, maintenance and support are needed — economic evaluations must include training, cyber‑security and support costs.
• Distributional and equity considerations:
  • Differential access to digital tools can produce asymmetric compliance burdens and potential regressivity (smaller firms and low‑income taxpayers may face higher relative costs or exclusion).
  • Policy must anticipate and mitigate the risk that automation exacerbates the digital divide and uneven enforcement.
• Design and governance of algorithmic systems:
  • Transparency, explainability and accessible appeal mechanisms are economically consequential: they affect perceptions of legitimacy and therefore voluntary compliance rates.
  • Economists should incorporate behavioral heterogeneity (age, digital literacy, sector) when modelling the impacts of automated enforcement and designing targeted nudges.
• Labour and organisational effects:
  • Automation will reallocate tasks within tax administrations (fewer routine processing roles, more data analytics, oversight and taxpayer support), with implications for public‑sector labour markets and retraining needs.
• Research agenda for AI economics:
  • Need for causal, microdata‑based studies (RCTs, natural experiments) to estimate elasticities of compliance to different automated interventions.
  • Evaluate long‑run effects on tax morale, administrative costs, shadow economy size and firm behaviour.
  • Study algorithmic fairness, distributional impacts, and how transparency/appeals affect compliance incentives.
  • Model macro‑fiscal implications of widespread real‑time reporting on revenue volatility, tax policy design and enforcement incentives.
• Policy recommendations implied:
  • Pair automation with investments in digital literacy and taxpayer support.
  • Build governance frameworks: algorithmic transparency, data protection, legal clarity, and user‑centric design.
  • Use segmentation to deploy differentiated interventions (simple interfaces and pre‑filled returns for low‑knowledge taxpayers; targeted nudges and monitoring for high‑risk firms).
  • Monitor and evaluate implementation outcomes to prevent exacerbating inequality and to refine cost‑effective enforcement.

If you want, I can convert these implications into a short policy brief for tax authorities or outline an empirical research design to measure causal effects of a specific automated intervention (e.g., e‑invoicing or AI risk alerts).