We’re in the middle of planning a migration from Camunda to open-source BPM, and the licensing situation is getting messy. Right now we’re running separate subscriptions for different AI tasks—OpenAI for some workflows, Claude for others, Deepseek for specific analysis work. It’s gotten to the point where finance can’t even tell us what we’re actually spending month to month.
When we started looking at open-source options, someone mentioned consolidating to a single subscription with access to 400+ models instead of managing all these separate API keys and contracts. The math seemed like it should work, but I’m stuck on how to actually build a business case that accounts for this.
The problem is that every time I try to model the cost savings, I either oversimplify it (“we’ll just save money”) or I get lost in a spreadsheet with too many variables. We need to show finance what consolidation actually saves, but also account for the migration effort, any retraining on new tooling, and whether we can actually migrate our existing workflows without rebuilding everything from scratch.
Has anyone actually built a realistic cost model when you’re consolidating multiple AI subscriptions alongside a platform migration? What variables did you use, and how did you handle the ones that were hard to quantify?
I went through this last year when we moved from managing separate Anthropic and OpenAI contracts to a unified approach. The honest thing is most people leave money on the table here.
What actually worked for us was breaking it into three parts. First, direct costs—just the monthly spend on each API. That’s the easy one. Second, operational overhead—how much time does your team spend managing keys, tracking usage across platforms, dealing with billing issues. We found that was closer to 2-3 headcount hours per week, and when you add that up it’s real money.
Third part was the migration cost itself. Don’t try to magic it away. We kept existing workflows running for six months in parallel while we migrated gradually. That meant paying for both systems during transition, but it meant we didn’t blow up production.
The number that surprised us was how much we were overpaying because we’d bought different tiers on different platforms. We had Claude Pro on one account, pay-as-you-go on another, and enterprise pricing on a third. A single subscription with everything available actually leveled out our cost curve across the board.
I’d suggest tracking your current spend by workflow type for the last three months. Then model out what each workflow would cost under unified pricing. The gap between those two numbers is your case.