Amazon has made Textract Insights generally available, adding generative extraction to its document-processing service. Users describe the information they need in natural language, then submit a document or a packet containing different document types. The service returns structured fields rather than requiring a separate template for every layout.

Each extracted value includes a confidence score and page-level grounding so reviewers can trace it back to the source. That design targets workflows such as applications, claims and contracts where a plausible answer is not enough and staff need to verify where a value came from. Mixed packets can be processed without first sorting every page into a rigid form category.

The release should reduce setup work when organizations receive changing document formats, but confidence and citations do not eliminate review requirements. Poor scans, ambiguous wording and unsupported field requests can still affect results, and AWS’s announcement does not substitute for testing on an organization’s own documents. Teams adopting the feature should set thresholds for human review and measure field-level accuracy before allowing extracted values to trigger financial, legal or customer-facing actions.