Top 5 Jobs in Government That Are Most at Risk from AI in San Marino - And How to Adapt

AI risk matters for San Marino because international rules and real-world uses are converging on the tiny republic: San Marino joined the Council of Europe Framework Convention on AI, which presses governments to bake in transparency, oversight and remedies, and UNECE work shows small states are already using AI in official statistics and smart-city planning - where mistakes can undermine rights or public trust.

At the same time, cross-border trends like the EU AI Act's risk categories and presumptions about high‑power GPAI models and the patchwork of global measures mean even local deployers must document, register and manage AI-enabled systems (EU AI Act regulatory tracker).

Practical applications such as fraud detection and identity verification use cases in government promise gains but raise privacy and bias risks, so upskilling civil servants with workplace AI training can turn threat into opportunity.

“Canada is proud to sign the first international convention on AI and human rights. This will reinforce human rights and democratic norms in AI governance, while strengthening transatlantic and global cooperation on AI.”

Assessment relied on a pragmatic, source-driven rubric tailored to San Marino's compact public service: roles were evaluated against four practical dimensions drawn from the literature - task routine and repeatability (the classic automation signal), data intensity and governance, rights‑impact or potential for biased outcomes, and workforce readiness for reskilling - so abstract risk becomes a clear, rankable score rather than rhetoric.

This approach followed Info‑Tech's recommendation to use an Automation Maturity Assessment Framework to map where automation can be safely scaled (Info‑Tech Automation Maturity Assessment Framework for public sector IT), folded in Guidehouse's emphasis on baseline skill assessment and tailored upskilling pathways (Guidehouse public-sector workforce readiness guidance for AI and intelligent automation), and kept the regulatory and governance caveats from The Regulatory Review in view - especially the need for transparency, monitoring and end‑of‑life plans for AI tools (The Regulatory Review chainsaw approach warning on public-sector automation).

The result: a defensible, repeatable method that flags high‑risk roles needing human oversight and targeted training rather than wholesale cuts.

“with 69% of survey respondents acknowledging this gap, organizations may struggle to deploy and scale Generative AI solutions effectively.”

Tax auditors at the Segreteria di Stato per le Finanze sit squarely at the intersection of opportunity and risk: many core duties - reconciling ledgers, authorizing journal entries, preparing periodic reports and supporting audits - are highly routine and data‑intensive (see the City of San Marino class specification), which makes parts of the job attractive for AI automation but also raises stakes for accuracy and transparency.

In San Marino the Central Bank's Board of Auditors already exercises strict oversight over accounting and the veracity of financial statements, so any algorithmic assistance must be paired with clear traceability and human review (Central Bank of San Marino Board of Auditors).

Practical adaptation is straightforward: keep humans in the loop for high‑risk judgments, invest in upskilling on anomaly detection and AI literacy, and pilot tools on reconciliation and fraud‑flagging with robust controls - Nucamp's government use‑case guidance on fraud detection shows how anomaly scoring can help while limiting false positives (Nucamp AI fraud detection use cases guide).

Remember: in a compact republic a single misposted journal entry can ripple through the budget like a dropped pebble in a pond, so preserving auditability and human oversight is non‑negotiable (City of San Marino class specification).

Customs officers at the Ufficio Dogane e Accise sit at a high‑stakes junction where routine, data‑heavy work - tariff classification, MRN tracking, paperwork and duty calculations - meets unpredictable enforcement: inspections, seizures and judicial investigations to stop counterfeits and protect public health.

Regional practice from the Agenzia delle Dogane e dei Monopoli shows the split clearly: digital services like MRN tracking and OTELLO 2.0 can speed clearances and target risky flows, but judicial functions and on‑site authenticity checks still require human expertise (Agenzia delle Dogane e dei Monopoli - customs information and digital services).

That duality is why automation should be applied as augmentation, not replacement: pilot AI for anomaly scoring, selective document pre‑checks and AEO‑linked streamlining, while preserving trained officers for inspections, seizures and cross‑agency investigations described in ADM's law‑enforcement guidance (Agenzia delle Dogane law‑enforcement guidance).

Upskilling in tariff coding, digital casework and AI‑aware evidence handling - plus adopting proven fraud‑detection patterns from government use cases - will let San Marino tighten borders without losing the human judgment that turns a red flag into a lawful seizure rather than a false positive (Nucamp AI Essentials for Work - fraud detection and identity verification syllabus).

A vivid reminder: ADM reported more than 2.2 million counterfeit items seized in 2016, with an assessed value exceeding €11 million - proof that mistakes or over‑automation can cost both money and trust.

Key 2016 metrics: Total counterfeit pieces seized - 2,232,449; Assessed value (euro) - €11,118,193; Estimated value (euro) - €16,568,106.

Administrative clerks in San Marino carry outsized responsibility for civic trust - preparing council agendas, certifying official records, running elections and responding to Public Records requests - so automation here is not just about speed but about legal soundness and permanence: the City Clerk's office explicitly

maintain[s] and publish the San Marino Municipal Code

and is custodian of official records, which means any tool that helps must preserve certifiable audit trails (San Marino City Clerk - municipal code and official records).

The detailed class specification for the clerk highlights routine, repeatable tasks well suited to augmentation - agenda packet assembly, minute‑taking, indexing and records retrieval - but also stresses strict deadlines, confidentiality and statutory duties that demand human oversight (San Marino administrative clerk class specifications - GovernmentJobs).

Practical adaptation: pilot AI for automated transcription, agenda formatting and search‑assisted records retrieval while keeping humans responsible for legal certifications, election filings and redactions; couple pilots with training on records‑management systems and documented review workflows so that a single misfiled ordinance doesn't become an invisible rule.

For guidance on designing low‑risk pilots that balance efficiency with privacy and false‑positive controls, reuse proven patterns from government use cases such as anomaly scoring and identity‑protection workflows (AI fraud detection and identity verification use cases for government).

Civil registry clerks at the Ufficio Anagrafe e Stato Civile will be on the front line of a quiet but profound shift as San Marino issues a new multifunctional, contactless national ID built on the HID SOMA platform: the same card will serve as residence permit, driver's license and even a health card, so enrollment accuracy matters in ways it didn't before.

That convergence speeds service delivery but raises clear privacy and data‑protection duties - biometric information is treated as a special category under EU rules, so clerks need documented consent workflows, short retention periods, and reliable encryption and audit trails to avoid “function creep” or wrongful reuse (see HID Global press release on San Marino's contactless national ID, Thales overview of biometric data and privacy laws, Nucamp Cybersecurity Fundamentals: guide to identity verification and fraud detection).

“Our search for the most advanced technology, combined with a commitment to adopting the highest international security standards, guided our partnership with HID and N.I.D.O. Group.”

Social services caseworkers at the Segreteria di Stato per la Sanità perform intensely human, high‑stakes work - assessing family needs, developing case plans, coordinating with health and education providers and responding to crises - so the upside of automating routine tasks must be balanced against privacy, cultural‑competence and crisis‑response risks highlighted in the field (caseworkers often handle home visits, documentation and heavy caseloads that demand both empathy and organization).

Practical adaptation for San Marino: pilot AI as an assistant for intake triage, appointment scheduling and search‑assisted resource matching while keeping humans responsible for risk assessments, emergency interventions and court testimony; pair pilots with mandatory training in digital case‑management tools, ethical decision‑making and data security because electronic records and confidentiality are central to social work practice (see guidance on what caseworkers do and the skills they need).

Protecting client privacy must be baked into every rollout - use privacy‑preserving patterns from Nucamp's government use cases for identity verification and anomaly scoring to reduce false positives - and require documented human review for any negative eligibility or safety decision.

The clearest warning is practical: when a family calls at midnight, an incorrect risk score that delays a home visit isn't an abstract error, it's a real harm that makes the case for “human‑in‑the‑loop” nonnegotiable; build training, audit trails and clear escalation paths before scaling.

“Case workers make sure their clients' needs are being met and will help clients access the resources and support necessary to improve their well‑being and quality of life,” said Dr. Thomas MacCarty.

San Marino's path forward is straightforward: treat AI like public infrastructure - first find it, then govern it, then train for it. Start with a rapid inventory and lightweight procurement review to catch “shadow AI” tools before they leak data or undermine sovereignty (see Forrester's guidance on shadow AI risks and faster review processes), align pilots with the 2020 national AI policy and local innovation partners to ensure interoperability and resilience (San Marino AI policy and innovation overview), and design all high‑risk deployments with clear human‑in‑the‑loop controls, documented audits and short retention for sensitive IDs and biometrics.

Parallel actions: mandate basic cyber hygiene and fast incident reporting, run small scenario‑planning exercises tailored to a microstate's cascading risks, and invest in targeted upskilling so staff move from fear to capability - for example, a pragmatic 15‑week course that teaches AI use at work, prompt design and job‑based deployments can get teams ready quickly (AI Essentials for Work bootcamp).

In a compact republic, small, governed steps now buy big trust later.