What the assessment week produces.
WORKED EXAMPLE FROM PUBLIC ACCOUNTS
Not a client engagement. Built from a listed manufacturer's published 2025 annual report to show the method. The company was not involved.
The situation
A listed European manufacturer holds about a year of stock. In 2025 its sales fell, yet finished goods in stock went up. Sales, plants and purchasing plan to different numbers, and finance only sees the cost at month-end.
The question
Should it add AI agents to its monthly sales and operations planning, and is it worth it?
The answer in three numbers
- €1.4m
- Pilot cost, approved now
- €4.0m
- NPV at 10% if the likely case holds; payback in 2.8 years
- €13.3m
- Stock turned back into cash at full scale, counted once
The honest part
In the conservative case the full rollout loses money. So the decision is a pilot with a gate, not a rollout. If the pilot misses the gate, it stops, and the loss is about €1.3m.
The gate, end of year one
- Forecast accuracy up 5 points
- Finished-goods stock days down 8.5%
- Write-downs falling
- Planners accepting the agents' proposals
Who owns it
The CFO owns the benefit and reviews it every quarter. Agents prepare the plan; people take every decision.
Rules applied
- Your numbers set the baseline; benchmarks only set the range.
- Cash released is counted once, never as a yearly saving.
- Running costs are in the case from year one.
- A benefit without a named owner is not counted.