I’m working through an enterprise automation decision right now and the financial side is honestly getting messy. We’re looking at Make vs Zapier, but the complication is that we also need multiple AI models integrated into our workflows—Claude, GPT-4, some specialized ones. Right now we’re paying subscriptions to like five different AI services separately.
What’s throwing me off is that I can’t find a straightforward way to compare total cost of ownership across these platforms when AI licensing is part of the equation. Make and Zapier pricing models are different enough, but when you add in the cost of running multiple AI model subscriptions on top of either platform, the math gets really opaque.
Some vendors are talking about consolidated AI pricing now. I’m trying to understand if that actually simplifies the comparison or if it’s just another variable to plug in.
Has anyone actually built out a proper TCO model that accounts for platform licensing, per-task costs, AND the cost of AI model access? How do you even structure that comparison to make it meaningful for leadership?
We went through this exact exercise last year. The issue is that most TCO models treat platform costs and AI costs as separate line items, but they actually interact.
What we did was build a model that starts with your actual workflow volume. Like, not theoretical—how many automations you’re actually running per month, how many API calls each one makes. Then we layered in the per-model costs for each platform.
The breakthrough came when we realized Zapier was charging us per action, and every time we needed Claude, that was another action. Make was cheaper on a per-task basis, but we were managing API keys for five different services. Once we accounted for the operational overhead—like security reviews, credential rotation, support time—the picture changed.
The big miss in most comparisons is that people don’t factor in the cost of integration complexity. If you’re creating a workflow that needs to swap between models, some platforms make that way easier than others, and easier usually means less developer time.
One thing I’d add: if you’re looking at unified AI pricing, be really careful about lock-in. We looked at a couple of options that promised to consolidate everything into one subscription, but the catch was that you lose flexibility if a newer model comes out that fits your use case better. You’re anchored to whatever’s in their bundle.
The TCO model that worked best for us separated recurring costs (platform + AI licensing) from variable costs (volume-based charges) from one-time costs (migration, training). That way you could scenario-play. Like, “if we go from 500 to 1000 workflows, how does cost scale?” And you could see which platform got more expensive faster.