Pode a IA auditar e apresentar declarações fiscais de forma autónoma para 10 milhões de pequenas empresas sem intervenção humana, integrando bases de dados contabilísticas e códigos fiscais ?

Os sistemas de IA estão cada vez mais a lidar com tarefas numéricas complexas, mas a conformidade fiscal requer nuances legais e responsabilização. À medida que as regulamentações se digitalizam e as auditorias se tornam algorítmicas, a etapa final poderá ser a automação total.

Background

AI systems have progressed in handling complex numerical tasks, yet tax compliance introduces legal nuance and accountability requirements. Digitalization of regulations and algorithmic audits have advanced, but full autonomy in tax filing remains out of reach.

Contemporary tax engines employ natural-language processing (NLP) to extract transaction data from unstructured accounting records and deep-learning models to classify expenses and deductions. For example, systems can interpret jurisdiction-specific tax rules and generate draft returns based on inputs from accounting databases. However, current implementations still rely on human review to ensure accuracy, address errors, and maintain compliance with rapidly evolving tax statutes.

Technically, real-time integration with a wide range of accounting platforms is feasible, but critical gaps persist. These include the absence of standardized audit trails, unclear responsibility allocation in the event of errors, and the lack of regulator-approved e-signature workflows that meet compliance standards. As a result, most commercial solutions provide "co-pilot" automation—assistive AI that augments human effort—rather than fully hands-off systems.

— Enriched May 10, 2026 · Source: OECD