INCREMENTAL ARTIFICIAL INTELLIGENCE GOVERNANCE IN A SMALL TECHNOLOGY-BASED COMPANY: A FRAMEWORK BASED ON CAPABILITIES AND DEPENDENCY RELATIONSHIPS
Abstract
The adoption of Artificial Intelligence (AI) systems by small businesses increases the need for mechanisms capable of controlling risks, defining responsibilities, ensuring traceability, and supporting human oversight. This study develops an incremental AI governance framework based on capabilities and dependency relationships through Design Science Research and a single case study conducted in a small Brazilian technology-based company that employs generative AI and multimodal processing for veterinary decision support. The investigation combined scientific literature, governance frameworks, source-code inspection, automated testing, architectural documentation, operational workflows, and decision records. A total of 31 governance mechanisms were identified and consolidated, classified as implemented, partial, or planned. Partial mechanisms were characterized according to gaps in coverage, integration, traceability, validation, and institutionalization, while relationships among capabilities were examined through a dependency matrix. The results revealed greater consolidation of mechanisms associated with day-to-day operations, such as identity, authorization, consent, prompt versioning, review, and curation, while gaps remain in end-to-end traceability, systematic validation, independent dual review, versioned datasets, and governed evolution. The resulting framework organizes capabilities into five levels and four cross-cutting elements, without assuming a rigid maturity path, providing a structure for the analysis and gradual evolution of AI governance in small organizations.