Field-Level Data Extraction

This is the reading of specific fields, like invoice numbers, dates, totals, or names, out of scanned or uploaded documents inside Cradl AI. A Beam agent triggers when a new document lands in a queue, reads the fields Cradl AI has pulled out, and applies the customer-approved mapping to move each value into the right business system, such as an accounting or records tool. Where a field comes back with a low confidence score or an unexpected format, the agent leaves the document in a queue and routes it to a person to confirm before anything downstream changes.

Document Routing and Filing

Cradl AI processes many kinds of internal paperwork, and each type usually needs to end up in a different folder, system, or approval chain. A Beam agent reads the document type and status Cradl AI assigns, then applies the filing rule a customer has approved, such as sending contracts to a legal drive and invoices to finance software, and updates the record to show it has been filed. When a document type is missing, mixed, or doesn't match any approved rule, the agent sends it to a person to sort and file by hand.

Low-Confidence Exception Queue

Not every document Cradl AI processes comes through clean: scans can be blurry, fields can be cut off, or handwriting can be hard to read. A Beam agent watches for documents Cradl AI flags below an agreed confidence threshold and checks whether a simple, pre-approved fix applies, like re-running a known template match. Where no approved fix resolves the issue, the agent holds the document in an exception queue, notifies the person responsible, and waits for their correction before the record moves forward in the process.