Artificial Intelligence Procurement and Vendor Dependency: Managing Lock-In Risk Among Small Businesses in Emerging European Markets
Abstract
Purpose: This paper examines how small businesses in emerging European markets procure artificial intelligence (AI) and what vendor dependency risks follow from those decisions. AI adoption remains low, at 8.4% of enterprises in Poland and 5.2% in Romania against an EU average of 20.0% in 2025, so procurement choices made now will shape dependency for years.
Design / approach: An integrative review combines transaction cost economics, resource dependence theory, switching-cost theory and Kraljic’s procurement portfolio model with research on cloud lock-in, Eurostat statistics and EU law. The analysis builds on Сидорук’s (2023) distinction between creating, ordering and subscribing to AI.
Findings : Purchased AI gives rise to eight forms of vendor dependency. Each implementation route carries a distinct dependency profile. The EU Data Act reduces infrastructure and data lockin, but model continuity, output drift and knowledge dependency remain matters of contract and management.
Practical implications : An AI dependency risk matrix, contract safeguards for each quadrant and a six-point procurement checklist help owner-managers match procurement strategy to the role an AI service plays in the business.
Originality / value: The paper links firm-level AI implementation routes to AI-specific lock-in risks and evaluates current EU regulation from the perspective of small buyers in emerging European markets.