Document Processing
A document already contains the details — they are just not where the system needs them. So today a person transfers them. Such an application should read the document, extract the required details and show which of them are recognised confidently and which need someone to confirm them.
A capability is a means. The linked areas describe what it is used for.
What specifically gets built.
Every line is a commitment about future work — none of it is a claim about completed projects.
The decisions behind it.
Categories rather than vendor names — which product is used depends on your existing systems.
- Evaluation by structure and meaning rather than fixed coordinates — so a project does not fail at every new sender layout
- A fixed output schema, so every result has the same shape
- A threshold per field type: below it, nothing continues automatically
- The original document stays untouched and remains the source
What is not included.
So that no time in the intro call is spent on an expectation I do not meet anyway.
- No percentages for accuracy — every field is marked either as recognised or as needing review
- No automatic posting without human approval
- No processing of documents for which no checking rule exists
These eight capabilities are means, not offerings. Which of them are used follows from the process — not the other way round. If you are not sure which ones you need, that is the normal case and not a problem.