Survey Says: You'll swap the model, but everything you build around it is staying put
Earlier this month, my team at Stacklok surveyed 520 technical leaders to understand whether they were concerned about being locked into frontier labs. The data says they’re already swapping models in and out, but the trap they’re concerned about is the layer on top of the model.
The recent Stacklok survey of 520 technical leaders reveals that enterprises are actively swapping AI models, yet remain concerned about being locked into the layer on top of the model. The survey indicates that 8% of respondents use just one model provider, while 25% utilize five or more providers. Companies with multiple providers are already working to build in optionality, with 76% having developed integrations to internal systems or MCP servers and 45% employing open-source agent frameworks.
The top concerns for moving between providers include integrations, multi-step agent workflows, guardrails, access control configuration, and delays in migration due to switching costs. Despite this, 84% of leaders prioritize avoiding lock-in in their AI strategy, and 80% would be interested in evaluating an open alternative to closed-stack architectures.
Organizations that already employ multiple model providers differ in their approach, with 67% considering portability across providers critically or very important, compared to 47% of those using only one or two providers. They are also more likely to consider adopting open-source alternatives to vertically-integrated stacks from frontier labs.
The survey highlights the need for organizations to prioritize flexibility as they increasingly run heterogeneous AI components, treating models as replaceable components behind stable interfaces.
Written by urgent.news from Stack Overflow Blog's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
