Most operators price AI work by the hour or by the token. The real number is cost per successful outcome. This prompt builds that number.
When you need to price a new AI-leveraged service, evaluate whether to keep automating a workflow, or renegotiate a retainer because the margin is gone.
Task description, model used, approximate cost per run (from your logs), success rate (percentage of runs that produce usable output), and current pricing (if any).
I'm building a pricing model for an AI-leveraged workflow.
Task: [describe what the AI does]
Model: [which model, input tokens per run, output tokens per run]
Cost per run: [from logs or estimation]
Success rate: [what percentage of runs produce usable output]
Current revenue per outcome: [what you charge now, or what you'd charge]
Calculate: (1) cost per successful outcome, (2) gross margin at current pricing, (3) break-even revenue per outcome.
Show me the math. If margin is negative, what would the revenue need to be?
Run this prompt against each workflow you're automating. Input real numbers from your logs or estimates if you don't have logs yet. The output tells you whether the workflow is viable at your current pricing or what you need to charge to make it work.
A clear table showing cost per run, success rate, cost per successful outcome, current revenue, and gross margin. A line stating break-even revenue. A note on which workflows are profitable and which are margin traps.
Using average cost instead of p95 cost. Assuming success rate is higher than it really is. Forgetting to include human review time in the cost.
Operators who do this once get rich. Operators who skip it automate themselves into a loss. Do this for every workflow you're considering. It takes 10 minutes and saves you months of guessing.