Switching from five separate AI subscriptions to one—what's the actual financial impact when you factor in integration work?

We’re currently paying for individual subscriptions to OpenAI, Anthropic, Cohere, and a couple others because different teams picked different tools over time, and nobody coordinated. Finance is finally asking questions about why we’re paying for overlapping services, and I’m tasked with figuring out if consolidating would actually save us money or if the integration effort would just eat any savings.

On paper, the cost consolidation looks good—we’d probably cut recurring spend by 30-40% if we moved everyone to a single platform with unified pricing. But what I can’t figure out is the hidden stuff. How much engineering time does it take to rip out the existing integrations and rebuild them in a new system? Do our workflows actually become more efficient, or are we just moving the same inefficiency to a different platform?

The other thing I’m worried about is whether the per-task costs actually work out better when you’re using a unified platform versus separate subscriptions. Like, if I’m paying one flat rate for access to multiple models, am I actually paying more per task than I was before, just spread differently?

Has anyone actually gone through a consolidation like this? What was the real time and resource cost to migrate, and how long did it take before the cost savings actually materialized? I need numbers I can take to finance.