What changed
Companies built gateways and routers so engineers could switch AI models by changing a single line of code. But a VB Intelligence finding says the change takes weeks for most companies, turning an apparently simple upgrade into a much longer job.
Why it matters
The promise of model gateways is flexibility: when a better model appears, a company should be able to route work to it quickly. The finding suggests that this flexibility is largely an interface promise, not yet an operational reality.
That matters because model providers keep leapfrogging one another in intelligence. A company may have the switch in its codebase, but changing models can still involve implementation work that stretches across weeks. The report does not specify which tasks cause the delay, so the practical bottleneck remains unclear.
The useful takeaway is straightforward: “one line of code” describes the control surface, not the full migration. Until the surrounding work becomes faster, companies cannot treat model choice as a truly on-demand decision.
Comments
No comments yet.