AI Costing Prompts · By task
Pressure-test a price change against the cost floor
Before you cut or raise a price, run the move through this prompt. It computes contribution per unit and in total for the current price and both scenarios, picks the option that maximises total contribution, and tells you exactly how much extra volume a price cut would need just to break even.
In short
Give the model the price, the variable product cost, the cost to serve per unit, current volume, and your assumed volume response to a 5% cut and a 5% rise. It works out contribution at each price, multiplies by the expected volume, and compares total contribution. It also calculates the break-even volume increase the price cut would require, which is usually far larger than people expect, and flags the assumption that quietly decides the answer.
What the prompt is doing
The core idea is that price decisions live or die on contribution, not on revenue or on margin percentage. Contribution is price minus the costs that genuinely vary with the unit, and the only fair comparison between scenarios is total contribution: contribution per unit multiplied by the volume you actually expect at that price. A price cut can grow revenue and still shrink the money you keep, because the lower margin per unit has to be made up by volume you may not get.
This is also where cost to serve has to be handled with care. If you treat cost to serve as fully variable, every scenario looks cleaner than reality, because some of that cost is fixed and will not fall when volume drops. Before acting on any pricing decision, the variable and fixed portions of cost to serve should be separated, otherwise the cost floor in your analysis is lower than the cost floor in your business.
The prompt
You are a pricing and cost analyst. Work only from the data I give you. Do not invent any numbers. Flag any assumptions clearly. One product: Current price: 100 EUR Variable product cost: 52 EUR Cost to serve per unit: 18 EUR Current monthly volume: 2,000 units If I cut the price 5%, volume rises 12%. If I raise the price 5%, volume falls 9%. Do the following: 1. Calculate current contribution per unit and total contribution. 2. Scenario A: cut the price 5%. Show new contribution per unit, new volume, and total contribution. 3. Scenario B: raise the price 5%. Show the same. 4. Compare and recommend the option with the highest total contribution. 5. Calculate the break-even volume change the price cut would need to keep total contribution flat. 6. List any assumptions.
A worked example
Current contribution = 100 - 52 - 18 = 30 EUR per unit, total 60,000 EUR at 2,000 units.
| Scenario | Price (EUR) | Contribution/unit (EUR) | Volume | Total contribution (EUR) |
|---|---|---|---|---|
| Current | 100 | 30 | 2,000 | 60,000 |
| A: cut 5% | 95 | 25 | 2,240 | 56,000 |
| B: raise 5% | 105 | 35 | 1,820 | 63,700 |
The price cut grows volume but loses money: total contribution falls to 56,000, down 6.7%. The price rise sheds some volume but each remaining unit earns more, lifting total contribution to 63,700, up 6.2%. The price rise wins.
Break-even for the cut: you would need 60,000 / 25 = 2,400 units, a +20% volume increase, just to hold contribution flat. Only +12% is expected, so the cut has to beat the forecast by 8 percentage points before it even stops losing money.
Assumption flagged: cost to serve is treated as fully variable here. In reality part of it is allocated fixed cost that will not fall with volume. Split the fixed and variable portions before deciding, because that split moves the break-even and could change the recommendation.
What it costs you to run
The input is around 270 tokens. A complete run with both scenarios, the recommendation and the break-even calculation costs a fraction of a cent on any current model. It is cheap enough to run for every product you are tempted to discount, which is exactly when the arithmetic is most worth doing.
Token figures are approximate and vary by model and tokeniser.
The guardrail that matters
Work only from the data I give you. Do not invent any numbers. Flag any assumptions clearly.
Pricing prompts are dangerous when a model "estimates" a cost or a price elasticity to fill a gap. This line forces it to use only your figures and to surface the one assumption, the fixed-versus-variable split of cost to serve, that actually decides the answer. See how to stop AI inventing numbers.
When you need the real model
This prompt is only as good as your cost floor, and most cost floors are wrong because cost to serve is averaged rather than traced. Getting the variable and fixed split right, by product and by customer, is the work that makes pricing decisions safe to act on.