Why average handling time misleads
A business case built on average handling time assumes the automated path and the human path do the same work. They do not. Automation absorbs the straightforward cases first, which raises the average difficulty of everything left for a human and pushes cost into the exception queue that the original model ignored.
This paper reframes the calculation around four cost components so that the saving is stated as a range with named assumptions, rather than a single number that quietly depends on the easiest cases.
The four components
- Handling cost — the marginal cost of processing a clean case end to end
- Exception cost — the cost of the cases the system defers, including investigation time
- Assurance cost — sampling, review and audit effort required to keep the automation trusted
- Change cost — the ongoing cost of document format drift, policy change and re-evaluation
Building the model
Start from a case-mix sample rather than a process map: classify a real month of documents by difficulty and by whether a decision required judgement. Cost each band separately, then apply an automation rate per band instead of one rate for the whole process.
The sensitivity that matters most is not the automation rate but the exception rate, because every deferred case carries both handling and assurance cost. A model that does not show the saving at a doubled exception rate is not decision-grade.
What good looks like
- Savings expressed as a range with the exception rate as the swing variable
- Assurance effort funded explicitly, not absorbed informally by the operations team
- A re-evaluation budget for document and policy drift in year two onwards
Licence and citation
Free to read, quote and cite with attribution to SJAIN Tech. Redistribution of the PDF in modified form is not permitted.
Cite as: SJAIN Tech (Sjain Ventures Ltd), “The automation economics of document-heavy workflows”, 21 July 2026. https://tech.sjain.io/research/automation-economics-of-document-workflows