Calculate automation ROI before you build anything
Most automation business cases count hours saved and ignore exceptions, change management and run costs. A simple model that finance will actually sign off.
'This will save 10,000 hours a year' is the most common line in automation proposals, and the least believed. Finance teams have seen too many projects where the hours were saved on paper and the headcount never moved.
A model finance will sign
We use four numbers per workflow, each measured from real data rather than estimated in a workshop:
- Volume: cases per month, from system logs, not memory.
- Handle time: median minutes per case, sampled by observation.
- Automatable share: the percentage of cases that fit the happy path. This is where optimism hides.
- Run cost: inference, infrastructure, licences and the human review that remains.
Where the real value usually hides
Hours saved is rarely the biggest number. Faster cycle times reduce working capital. Fewer errors reduce write-offs and rework. Same-day responses improve retention. In lending, for example, the headline is often conversion rather than underwriter hours: applicants who get a same-day decision have less reason to go to a competitor.
- Cycle time → cash flow and customer conversion
- Error rate → write-offs, rework and compliance risk
- Capacity → growth without proportional hiring
The takeaway
Measure first, model honestly, and include the exception tax. A smaller, believable number gets funded. A big, fragile one gets a 'let's revisit next quarter'.
- Automation
- ROI
- Operations
