The setup variables
Number of document types, heterogeneity of formats, number of systems to connect and quality of their APIs, complexity of the routing rules, number of human validation points, security and privacy requirements.
The more heterogeneous the formats and the more numerous the rules, the more the design and testing phase weighs.
The run variables
Monthly volume processed, provider costs (models, transcription, OCR), exception rate and the associated human time, monitoring and support, changes to the process.
A system with a high exception rate is expensive in human time even if the technology is cheap.
The assumptions to write down
Current handling time per document, fully loaded hourly cost, share of documents that can actually be automated, current error rate and its cost. Any savings estimate must show them.
We refuse to announce a saving without these assumptions. An honest calculation shows its variables, never a certainty.
What we do in Discover
Measure on your real documents, test extraction on a sample, map the integrations, and decide between accelerator, hybrid or Custom. That is when a quote becomes possible.
Illustrative example of how the process can work.