The numbers from our own projects
A UK wholesale distributor: £12,500 build, £42,000 a year saved, order-to-delivery from 3.5 days to 18 hours, order errors down 93%. Payback around 4.5 months.
An HR SaaS AI support layer: under 5 months. A call-centre voice deployment: $14,500+ a month net savings against a build in the tens of thousands.
These were chosen because they were good candidates. Yours may be slower, and the way to find out is to measure rather than to assume either way.
What delays payback
- Human review that never falls — if thresholds are never tuned, you keep paying for checking
- Capacity nobody uses — freed hours that do not become revenue or absorbed growth
- Running costs underestimated: API tiers, licences, maintenance
- Adoption failure — the parallel spreadsheet that never goes away
Count only what you can point at
If you cannot name the invoice that stopped or the hire that did not happen, it is a soft benefit. Soft benefits are real and they belong in a clearly labelled section of the paper.
- Time redeployed — only if you can say what the hours went to
- Errors avoided, priced from real historical error costs
- Volume handled without hiring
- Revenue attributable to faster response, where you can show the link
Take the baseline first
Two weeks of measurement before anything changes: minutes per unit, errors per hundred, elapsed time start to finish. It costs almost nothing and it converts every future argument into arithmetic.
We ask clients to do this whoever ends up building the system, including when it is not us.
Model it twice
Once with your realistic numbers, once pessimistically — 60% of the time saving, half the error reduction, top of the running cost range. If the pessimistic case still pays back inside two years, it is a sound project.
If only the optimistic case works, it is a bet rather than an investment, and it should be scoped smaller.