Start with the whole process
I would first ask someone to show me the work from beginning to end. Where does information arrive? Who checks it? What happens when a document is incomplete or a figure is wrong? The slowest visible step may not be the real constraint.
A sensible proposal explains what changes for the business. That may be more capacity at a busy time, fewer corrections or a faster response to customers. Each needs a baseline so the result can be assessed.
A worked invoice example
This is a fictional calculation, not a client result or a forecast. Assume a finance team handles 400 invoices a month. Entry and checking take six minutes per invoice. A trial reduces that combined effort to two minutes.
| Measure | Calculation |
|---|---|
| Current work | 400 × 6 minutes = 40 hours |
| Work after the trial | 400 × 2 minutes = 13.3 hours |
| Capacity released | 26.7 hours a month |
| Illustrative value at £25 an hour | About £667 a month |
| Assumed running costs | £200 a month |
| Net capacity value | About £467 a month |
If introduction costs £3,000, that implies roughly 6.4 months to recover the cost on these assumptions. The calculation only holds if the checking time is realistic, the released capacity is useful and additional maintenance does not consume the benefit.
Time released is not automatically money saved
If salaries stay the same, the business has not reduced its payroll by £667. It has released staff capacity. Ask what that capacity will be used for: resolving supplier queries, avoiding overtime, handling growth or something else with a clear value.
I would also compare the proposal with changes to the existing accounts system, better invoice intake or a simpler approval process. Include setup, licences, supervision, exceptions and maintenance in every option.
What I would want a trial to establish
Use a representative mix of documents, including the awkward ones. Measure the total time people spend, the errors that reach the next step and the cases that need manual attention. A fast demonstration using a few tidy invoices does not establish the business case.
Agree who checks the output, which information can be used and who can stop the process. Payment approval remains a separate business decision. The trial should make it possible to recommend proceeding, changing the design or stopping.
The decision I would put in front of you
A short assessment should set out the evidence, the options, the costs and the conditions on which the recommendation depends. Where a figure is still an assumption, I would identify what we need to measure before relying on it.