Document intake
Incoming documents are read, checked field by field and filed where they belong, with a person reviewing anything uncertain.
A pipeline that turns incoming documents (PDFs, scans, email attachments, forms) into checked data your systems can use. Each field is traced to its source, and low-confidence fields go to a person before anything is filed.
No list price. It starts with a diagnostic at a fixed fee, and the build is quoted after it.
Give it a document
Use a sample or one of your own. It is read, checked and sent on its way while you watch.
An image or a PDF, up to 3 MB.
Your file goes to the model once to be read. We do not store it. Do not upload anything confidential.
How it works
The pipeline, step by step.
- 01
Documents arrive
From an inbox, a shared folder, a portal or an upload.
- 02
Each document is classified
A declarations page, a bank statement, a signed engagement letter, a supplier invoice: the system knows which is which.
- 03
Text is recognised
Scans and photos are read, with the source image kept beside the result.
- 04
Fields are extracted
The fields you need, each with a confidence score and a link back to where it was found.
- 05
Rules are applied
Formats, required fields and cross-checks between fields, written with you.
- 06
A person reviews the uncertain
Low-confidence fields and rule failures go to a review queue. Everything else passes.
- 07
The record is written
Into your practice, policy or accounting software, or as a ready-to-enter record.
- 08
An audit log is kept
Who read, extracted, reviewed and filed each document, and when.
Connects to tools such as
- Gmail
- Outlook
- Google Drive
- SharePoint
- QuickBooks
- Xero
These are examples. We confirm the method for each of your tools before we build.
Where a person approves
The points where the system stops and waits for a person.
- You set the confidence threshold below which a person must check a field.
- Critical fields (an amount, a date, a name) are always shown to a person before filing.
- New document types are added only after you confirm the fields and the rules.
How it is measured
Named measures, scored on your own files before launch and after every change.
- Field-level accuracy on a locked test set of your own documents, with critical fields scored separately.
- Review rate: the share of documents that needed a person, and which fields caused it.
- Correction rate: how often a reviewer changed an extracted value.
How it is run
After launch: monitoring, updates and what you own.
- Monitoring on every document: failures, stuck items and unusual volumes raise an alert.
- Every change is re-scored against the test set before release.
- Handover: a runbook, a data map and code owned by the business.
What it will not do
Limits we set on purpose.
- It does not guess a value it cannot find. The field is left for a person.
- It does not overwrite a record in your system without a trace.
- It does not file a document whose type it could not determine.
Sectors we build for
The business types this system is built for.
Questions
Which documents can it read?
The ones you receive: forms, statements, declarations pages, invoices, signed letters, scans and photos. We test on your own files before launch.
How do we know the extracted data is right?
Every field carries a confidence score and a link to its source. The system is scored on a test set of your documents before launch and after every change.
What does a reviewer see?
The document and the extracted fields side by side, with the uncertain ones marked. They confirm, correct or reject, and the decision is logged.
Where does the data go?
Into the software you already use, or into a ready-to-enter record if you prefer to keep that step with a person.
Talk to us about document intake
Fifteen minutes on a call is enough to tell whether it fits your work. If it does not, we say so.