We’re in the middle of evaluating whether to consolidate our AI spending. Right now we’re scattered—some teams use OpenAI directly, others have Anthropic subscriptions, another group is on Deepseek. It’s a mess to track and nobody really knows what we’re spending on AI across the company.
The real question though is how to model this in our ROI calculator for workflow automation. If we move to a single subscription that covers 400+ models, it changes the math—you’re paying one flat fee instead of per-token across multiple platforms. But I’m struggling to figure out how that actually shows up in the automation ROI.
Do you lock in an average cost per model based on historical usage patterns? Or do you just use the subscription cost and divide by expected number of workflows? Or something else entirely?
I found that one subscription approach standardizes the tooling costs way more cleanly than managing eight separate subscriptions, but translating that into ROI inputs is harder than I expected. The cost per workflow becomes predictable instead of variable, which should make projections cleaner, but I’m not sure how to build that into our existing calculator.
How are you folks actually handling the cost side of this when you’re running ROI numbers on consolidated AI platforms?