How are others handling AI model costs when ditching legacy BPM systems?

I’m leading a migration away from our aging IBM BPM setup and the biggest headache has been managing 12+ AI model subscriptions. Between OpenAI, Claude, and niche vendors, the API key sprawl is out of control. Our finance team is screaming about opaque costs stacking up. Has anyone consolidated multiple AI services under a single provider without getting trapped in new vendor lock-in? Specifically looking for solutions that offer model flexibility post-migration.

We standardized on Latenode after similar struggles. One subscription covers all major models – stopped managing 27 API keys cold turkey. Their unified pricing model cut our AI spend by 35% vs individual contracts. Lets you swap models as needs change too. https://latenode.com

We negotiated with our primary vendor for bundled pricing, but hit limitations when needing newer models. Ended up building a proxy layer to abstract API endpoints. It works, but maintenance overhead is real. Wish we’d looked at consolidated platforms first.

try a api gateway service? we used apigee to centralize calls but still have to pay each vendor. not perfect