Skip to main content

How do I automatically classify incoming documents?

Configure extraction, classification, and metadata mapping as a processing pipeline. FormKiQ can use Textract-based processing and mappings, or a separately selected AI module, to identify document types and propose attributes. Validate results before triggering business decisions. Classification becomes useful automation when supported rules and workflows consume the resulting metadata and handle uncertain or failed results.

Define the classification contract​

Choose a small set of document types and the fields each requires. Keep example documents for ambiguous, unsupported, and mixed-content cases. Decide when an item needs human review instead of forcing it into the nearest class.

Configure OCR where the source needs text extraction. Use the installed IDP classification or extraction operations and the agreed mapping to document attributes. For model-backed classification, configure a named prompt and selected model before invoking analysis. Stored analysis output and authoritative metadata are separate until a mapping or approved update connects them.

Connect classification to routing​

After validating the result, use the resulting attributes in a ruleset or workflow. Test an invoice, a contract, an unsupported file, a low-quality scan, and a changed document revision. Confirm that processing failures do not accidentally reach an approval queue as successful classifications.

Record the source revision and extraction/prompt configuration when decisions need to be reproduced. Treat document content as untrusted input, including instructions embedded inside it.

Edition and configuration​

Textract OCR/IDP, mappings, rulesets, and workflows are included from Essentials with configuration. Generative AI Processing and Analysis is a separately selected Advanced/Enterprise extension.

Continue with the detailed guides​

Privacy settings

Analytics is optional. Your analytics choice is shared with formkiq.com and remembered for up to six months. Change it here at any time.

Essential preferences and search

We remember your privacy choice and display preferences. Search runs in your browser using Pagefind files hosted with the docs. It works whether you accept or reject analytics.

Measure page views and installation-link clicks. This does not enable advertising on the docs. Existing marketing choices for the main website are kept unchanged.

Read the docs privacy notice or manage other main-website privacy choices.