THE ROLE OF ARTIFICIAL INTELLIGENCE IN SOFTWARE-DEVELOPMENT PROJECT MANAGEMENT WITHIN INTERNATIONAL MEDIUM-SIZED ENTERPRISES (SMES): A SYSTEMATIC REVIEW

Abstract

Software-intensive small and medium-sized enterprises (SMEs) are adopting artificial intelligence (AI) to improve software-development project management (SDPM), from planning/forecasting and risk control to quality assurance/testing and developer productivity. To synthesize recent evidence (2023–2025) on (i) where AI is applied across the SDPM lifecycle in internationally active SMEs, (ii) the effects on delivery and quality outcomes, and (iii) governance practices that enable auditable, cross-border adoption.

We conducted a secondary-study systematic review (PRISMA-informed). Sources included peer-reviewed venues and authoritative grey literature (standards, regulatory texts, and programmatic guidance). Searches spanned academic databases and official portals for ISO/IEC 42001 (AI management systems), the NIST AI Risk Management Framework including the 2024 Generative AI Profile, PRINCE2 7, and the EU AI Act. Eligibility targeted SMEs (~10–250 employees) engaged in software development with international activity; interventions involved AI uses that influence SDPM; outcomes included DORA metrics (lead time, deployment frequency, change-fail rate, MTTR), schedule/cost variance, and defect/escape rates. From 1,100 records identified, 980 remained after deduplication; 110 full texts were assessed; 22 studies met inclusion criteria for qualitative synthesis.

Convergent evidence shows AI-augmented coding and testing deliver the most consistent near-term benefits (faster task completion, reduced defect leakage). Delivery performance improves when AI is layered onto strong engineering practices (continuous integration, automated testing, trunk-based development). Predictive planning and risk triage yield earlier variance signals in SMEs that implement minimum viable telemetry. Governance stacks built on ISO/IEC 42001, NIST AI RMF, and PRINCE2 7 support auditability, while the EU AI Act shapes provider/deployer obligations for international SMEs. Persistent barriers include skills and enablement gaps, data/telemetry immaturity, privacy/security risks (e.g., prompt leakage), and vendor lock-in.

AI can improve SDPM outcomes in internationally active SMEs when integrated with disciplined delivery and right-sized governance. We propose a three-layer framework—work level (co-pilots with review gates), flow level (DevOps analytics, test-impact and flakiness intelligence), and system level (AIMS/RMF/PRINCE2 governance)—plus a staged roadmap tied to measurable project metrics.