Small and mid-sized Class II medical device manufacturers face a persistent operating problem: regulatory documentation, supplier volatility, and market-access requirements often expand faster than compliance teams can scale. This study develops and evaluates an auditable AI-enabled operating model that connects regulatory readiness and supply chain resilience in Class II diabetes device operations. The research is based on a six-month pilot at New Horizon Biotech, where a five-FTE AI business unit supported FDA 510(k) and EU MDR workflows through retrieval-augmented generation, supplier early-warning scoring, human-in-the-loop review, and controlled evidence mapping. Baseline and pilot performance were compared across regulatory throughput, supply recovery, documentation quality, groundedness, and financial return. Results show that AI-assisted regulatory drafting reduced cycle time from 6.5 to 4.4 weeks, decreased defect density from 2.8 to 0.9 errors per page, reduced review iterations from 4.2 to 2.5 cycles, and shortened evidence retrieval time from 12.5 to 1.5 hours. Supply-chain outcomes also improved: lead-time recovery decreased from 8.5 to 5.2 weeks, sole-source exposure declined from 42% to 34% of the bill of materials, and critical stock-outs were avoided during the pilot window. A total-landed-cost simulation identified an 18.5% tariff threshold at which Mexico-based manufacturing becomes cost-neutral relative to the existing offshore manufacturing base. The study contributes a practical AI framework for regulated manufacturing without replacing quality-system controls or expert judgment.