According to the latest IBM Institute for Business Value (IBV) study released by IBM Japan, while more companies are incorporating AI into core operations, insufficient control, vendor dependence and data sovereignty challenges are increasing corporate risk.
A survey of 1,000 executives worldwide found that approximately 70% said it was difficult to switch major AI vendors or AI models, highlighting operational constraints. The same trend was observed in the Japanese market, where respondents pointed to the complexity involved when data is transferred and operated across national borders.
Meeting data residency and data sovereignty requirements was also cited as a challenge, increasing the burden on companies when moving and operating AI systems and data across different environments. As AI adoption expands and regulatory requirements become more stringent, the study underscored the need to strengthen control over AI further.
Insufficient visibility also emerged as a major challenge. Nearly 90% of respondents said they do not have sufficient visibility into dependencies spanning the AI vendors, AI models and infrastructure used by their organizations. As a result, assessing risks and developing contingency plans for disruptions has become more difficult.
Respondents also reported experiencing an average of 6 AI-related failures or disruptions over the past 2 years, with problems attributable to AI vendor services cited as a cause. Approximately 80% said that a seven-day outage of an AI vendor's services would have a serious impact on their operations.
The study suggests that companies building AI systems with the flexibility to adapt across data, models and infrastructure achieve better results than their peers. Although the proportion of companies using multiple vendors is also high, the analysis indicates that this is largely driven by operational requirements.