AI vendor lock-in raises migration costs and procurement risks

The Register published an opinion piece on 2026-04-28 arguing that enterprise AI vendor lock-in is becoming a material cost and operational issue. The article cites a Zapier survey of 542 US executives with active AI vendor contracts that found 41 percent said they could switch AI vendors in 2-5 business days, and that only 42 percent of organizations that attempted a migration reported it went smoothly, according to Zapier. The Register quotes Zapier: "The problem is that when AI is already woven into internal processes, connected to other systems, and tuned to specific workflows, it has dependencies, edge cases, and little adaptations that nobody documented because they were 'temporary.'"

Industry-pattern observations: Early adopters commonly embed models via vendor-specific APIs, proprietary fine-tuning data, custom deployment tooling, and bespoke workflow glue. These layers create technical debt that raises the cost and time of migrating between providers. Models referenced in the piece-Gemini 3.1 Pro, Claude 4.6, and GPT-5.5-illustrate the rapid churn at the frontier that enterprises confront when attempting to swap underlying model providers.

Public reporting frames this as the intersection of two trends: vendors moving away from loss-leader pricing and enterprises operationalizing AI in production. The combination increases the financial and project-management stakes of any migration. For procurement and platform teams, the consequence is that vendor selection, contract terms, and integration architecture now carry longer-term implications than many C-suite narratives suggested, per The Register's analysis and the Zapier data cited.

• •Contractual terms around data portability, SLAs, and exit clauses in AI vendor agreements
• •Adoption of abstraction or orchestration layers that aim to decouple business logic from vendor APIs
• •Survey follow-ups or vendor disclosures that quantify migration effort and costs

The Register reproduces Zapier's formulation: "The problem is that when AI is already woven into internal processes...it has dependencies, edge cases, and little adaptations that nobody documented because they were 'temporary.'"

the piece is an opinion column; the primary empirical input is the Zapier survey cited. The article frames the issue broadly but does not provide vendor-side statements or independent migration cost benchmarks.

Industry observers should track contract language and architectural choices that affect portability, and treat vendor migration as a multi-team project rather than a short swap.