INTEGRATING ARTIFICIAL INTELLIGENCE IN QUALITY AUDITS OF ORAL SOLID DOSAGE MANUFACTURING FACILITIES

In pharmaceutical manufacturing, traditional good manufacturing practice (GMP) audits are time-consuming, labour-intensive, and prone to subjective interpretations by inspectors. This may create inconsistent severity assignment, recurrent non-compliance, and limited capacity for cross-facility trend analysis. A harmonized and data-driven audit platform is required, where good practice (GxP) principles are maintained and predictive oversight is possible. To refine and test WinAI, an artificial intelligence (AI)-enabled audit platform implemented as a web-based system for oral solid dosage (OSD) manufacturing plants, the system incorporates harmonized checklists, natural language processing (NLP), a deterministic rule-based scoring engine, and historical-inference ability to (i) eliminate inspector bias in severity assignment, (ii) indicate repeated non-conformances, and (iii) provide real-time auditable dashboards to management and regulators. The WinAI architecture usually consists of a front-end audit user interface (UI) with a back-end rule engine, a natural language processing (NLP) pipeline and interpreters, and a recurrence-detecting historical database. Severity scoring is based on predefined rule sets (critical, major, and minor) and weighted aggregation across six good manufacturing practice (GMP) systems. Pilot validation involved simulated and limited live audit datasets, where the recorded measures included classification accuracy, audit duration, and inter-rater consistency. Pilot validation was performed preliminarily on 254 simulated and 12 live OSD audit results, in which expert auditors were used as the reference standard. The platform was preliminarily pilot tested, showing that WinAI can attain high accuracy in classifying good manufacturing practice (GMP) deviations (95.3 % accuracy; macro-F1 score of 94.1 %) and reduce the audit period by approximately 40 % compared to traditional practices. The recurrence-detection logic was able to identify repetitive non-conformances and auto-escalate the severity of recurring problems in favour of corrective and preventive action (CAPA) activity. WinAI offers an auditable mechanism for harmonizing good manufacturing practice (GMP) audits, lowering the level of inspector bias, and placing greater emphasis on systemic non-conformances. The platform is designed to meet good practice (GxP) validation expectations and support phased integration with regulatory inspection databases.