Data Overview
Box Data Extract surfaces content from your Box documents in four different ways, depending on how much processing you want Box AI to apply. Each entity targets the same folder but returns a different representation of the documents inside it.
Entities
Stream text representation folders
Plain text extracted from each document
Stream AI ask folders
AI-generated answer to your custom prompt, per document
Stream AI extract folders
Key-value metadata extracted by Box AI based on your prompt
Stream AI extract structured folders
Structured fields extracted by Box AI using a predefined schema
Fields by entity
Stream text representation folders
file_id
Unique identifier of the Box file
file_name
Name of the file
text_content
Full extracted text from the document
folder_id
ID of the parent folder
Stream AI ask folders
file_id
Unique identifier of the Box file
file_name
Name of the file
prompt
The question sent to Box AI
answer
The AI-generated response for that document
folder_id
ID of the parent folder
Stream AI extract folders
file_id
Unique identifier of the Box file
file_name
Name of the file
prompt
The extraction prompt used
extracted_data
Key-value pairs extracted by Box AI
folder_id
ID of the parent folder
Stream AI extract structured folders
file_id
Unique identifier of the Box file
file_name
Name of the file
structured_fields
Extracted fields matching the predefined schema
folder_id
ID of the parent folder
Common field combinations
file_name + text_content — useful for full-text search pipelines or feeding content into AI destinations
file_name + answer — great for summarization or Q&A dashboards built from document libraries
file_name + structured_fields — ideal for turning contracts, invoices, or forms into structured database records
Use cases by role
Extract text from policy documents or SOPs and load into a searchable Google Sheet
Use AI extract to pull key dates and parties from contracts into BigQuery for tracking
Feed structured extraction results into an AI tool like Claude or ChatGPT for further processing
Run structured extraction on NDAs or agreements to capture counterparty, date, and jurisdiction fields
Ask Box AI a compliance question across all files in a folder and collect the answers as a dataset
Append results from multiple folder extractions to build a unified contract register
Extract invoice data from scanned PDFs stored in Box using the structured AI entity
Pull vendor names, amounts, and due dates into a spreadsheet automatically
Use the Aggregate transformation in Coupler.io to summarize extracted invoice totals by vendor or period
Platform-specific notes
Box AI features (AI ask, AI extract) require a Box AI-enabled plan — not all Box accounts include this
The Folder ID is a numeric string visible in the Box URL when you open a folder (e.g.,
/folder/123456789)Enabling Recursive will process all subfolders, which can significantly increase the number of API calls and processing time for large folder trees
Each document in the folder is processed individually — very large files or folders with hundreds of files may take longer to sync
Supported file types for text extraction depend on Box's representation API; PDFs, Word documents, and plain text files are typically supported
Last updated
Was this helpful?