Skip to content
Understand
Tool

AI budget stress test

Does the case survive higher prices and more demand? Runs in your browser. Nothing is stored or transmitted.

Test whether an AI case still clears its hurdle if prices rise, demand grows, or both. Showing illustrative defaults. Replace them with your own case.

The case today

All AI cost for the case: usage, seats, platform and people.

Token and usage charges. The rest is treated as fixed.

Use measured value where you have it, and say which figure you used.

The return the case must keep. 1.5 means £1.50 of value for each £1.

Scenarios to test

2 tests a doubling. Below 1 tests a fall.

4 tests demand doubling, then doubling again. Below 1 tests a fall, which loses value at today's full rate.

100% if new use is as valuable as today's. Lower if growth goes to less valuable work.

The case misses its hurdle in 2 of 4 scenarios. At today's demand, it keeps meeting the hurdle up to a price rise of about 1.86×.

Value per £1 of cost, against a hurdle of £1.50

  • Today£2.40 Meets

    Cost £50,000, value £120,000 a month

  • Price ×2£1.41 Misses

    Cost £85,000, value £120,000 a month

  • Demand ×4£1.94 Meets

    Cost £155,000, value £300,000 a month

  • Both together£1.02 Misses

    Cost £295,000, value £300,000 a month

Hurdle

Usage cost moves with price and demand together; the rest is held flat, which flatters cases with large platform or people costs that also grow with use. A case that misses a scenario is not necessarily wrong. It needs a redesign, more commercial flexibility, or an explicit statement that its justification depends on today's prices. The hardest risk is not modelled here: a capability becoming embedded in core work before prices change, which limits how the organisation can respond.

The scenarios come from How to Budget for AI and Why Cheaper AI Will Cost More, which explains why demand tends to outgrow price cuts. For the cost side of the case, start with the inference cost estimator.