When a company chooses an AI provider, it considers price, performance, security, integration and compliance.
But I think one more question belongs on that list:
What will it cost us to leave the provider we choose today?
What is AI vendor lock-in?
AI vendor lock-in occurs when data, models, applications or workflows become so closely tied to one provider that moving elsewhere becomes difficult and expensive.
A 2026 study by the IBM Institute for Business Value surveyed 1,000 senior executives across 16 geographies and 17 industries. Among those who had switched or attempted to switch AI providers in the previous two years, 75% said the process was difficult.
Executives estimate that moving AI training and operational data to another environment would take an average of 145 days. And 57% believe replacing a core AI model would require significant decoupling or a complete system rebuild.
Is switching cost only a technical issue?
No. The real cost can include data migration, rebuilding integrations, model revalidation, compliance reviews, employee workflows and service disruption.
The industry is preparing for this risk. AWS describes LLM gateway architectures that provide a common interface across different models and providers. Microsoft recommends abstraction layers and parallel model testing to reduce vendor lock-in.
These approaches do not make switching free. But they can reduce the connection between an application and its model, helping businesses preserve choice.
Why does this matter for Ireland?
Ireland’s National Digital and AI Strategy, published in February 2026, contains 90 deliverables covering public services, skills and business AI adoption.
According to the latest Central Statistics Office data, 20.2% of Irish enterprises used AI in 2025, up from 15.2% in 2024.
To me, these figures point in the same direction. Adoption is accelerating, but many businesses are still building their AI architecture.
Companies should ask:
- Can our data be exported in a usable format?
- Can we change the model without rebuilding the application?
- Does the contract define exit support, data deletion and migration fees?
- Do we regularly test an alternative provider?
When planning how to adopt AI, we should also plan how to leave it.
Because the cheapest AI contract may not always produce the lowest-cost AI strategy.
