Building an Auditable Artificial Intelligence Layer in an Agri-Food SME: A Business Case Study of Universal Organic Products Ltd

Abstract

Small and medium-sized enterprises (SMEs) have been rapidly embracing artificial intelligence (AI), but most adoption remains limited to fleeting applications and is not well mapped. Most dot-com – can be now artificial intelligence (AI) adoption among small and medium-sized corporations (SMEs) is minimal and shallow, and was even less well-documented. This paper describes a single case study of a UK start-up marketplace for organic products, Universal Organic Products Ltd (UOP), which in 2026 adopted an integrated AI and automation layer and began producing a formal evidence pack for external review. Based on document analysis of technical explanation documents, workflow documentation, and platform screenshots, together with audit-layer counts of platform records, the study analyzed the design, execution, and proof of how eleven functions – ranging from marketplace assistant and allergy-aware recommendations to supplier matching and fraud review – were enabled by the use of AI. The common theme is that rather than any specific skill, UOP's most unique feature is an evidence-first architecture that ensures all AI decisions are saved into specially created audit tables, that data on the original platform is explicitly separated from content generated by the LLM, and that the data is not able to be backdated. This design serves three purposes: regulatory traceability, investor scrutiny, and organizational learning. The paper provides a grounded description of AI adoption in response to disparate evidentiary pressures, suggests evidence-first AI adoption as a dynamic capability for SMEs, and highlights risks associated with toxic products such as special-category health data, as well as allergen liability and cold-start marketplace risks.