AI unit economics
The cost of one unit of work your product does against the price you charge for it, then per customer and at 10, 50 and 200 customers.
Define your unit of work (one document read, one application processed, one ticket resolved), your pricing, usage and model prices. It shows whether gross margin holds as usage grows, and a sensitivity grid shows margin across model prices and usage.
What’s inside
- Inputs: unit, pricing model, usage, tokens, model prices and other costs
- Per unit: cost and margin for one unit of work
- Per customer and at scale: per customer a month, and at 10, 50 and 200 customers
- Sensitivity: gross margin across model prices and usage
Start here. Enter the price per unit, units per customer a month, tokens per unit, the two model prices and support cost per customer. Then read gross margin per unit and per customer, and check the Sensitivity grid.
Every file comes filled in with a worked example for Northlight AI Ltd, a fictional company. Blue cells are yours to overwrite; white cells calculate. Formats show £: for €, change the currency format on the number cells.
