Shipping Safer LLM Features in SMEs: A OnePage Checklist Mapped to NIST AI RMF and ISO/IEC 23894/42001
Small and medium-sized enterprises (SMEs) are deploying large language model (LLM) features without the governance capacity of larger firms. This paper synthesizes the NIST AI Risk Management Framework (AI RMF 1.0) and its 2024 Generative AI Profile with ISO/IEC 23894:2023 and ISO/IEC 42001:2023 to produce a concise, citable one-page checklist mapped to auditable clauses. The artifact was developed under a design-science method using a structured, rulegoverned mapping protocol applied to the two standards and to the official National Institute of Standards and Technology - NIST crosswalks. Organized by the RMF functions (Govern, Map, Measure, Manage), the result is a set of sixteen minimum-viable controls, each mapped to specific ISO/IEC clauses, together with a gap analysis contrasting typical SME practice with framework expectations, a framework crosswalk matrix, and a risk-lifecycle role allocation. The artifact further contributes a worked example of release-gate thresholds, a procedure for deriving local gate values from a measured baseline, and a lightweight eight-to-twelve-week validation plan suited to resource-constrained organizations. The example threshold values are illustrative and have not been empirically validated. We conclude that the distance between voluntary framework guidance and auditable management-system requirements can be closed for SMEs by a small, clause-mapped control set with explicit evidence requirements, and that the principal remaining barrier is threshold calibration rather than control selection. The significance is practical: SMEs obtain a short and defensible route from RMF guidance toward ISO/IEC 42001 certification practices. Empirical validation across multiple SMEs is planned as future work.