AI news

AWS extracts mixed files with confidence scores and page sources

In brief

According to AWS, Amazon Textract Insights is now generally available. You describe in plain language the fields to pull from a document or from a file that mixes several papers. The service returns structured values, each with a confidence score and the source page, without you training a model.

Source : AWS Machine Learning Blog - « Announcing Amazon Textract Insights: Verifiable generative AI for documents »

What it changes for an SME

If you already run on AWS and handle mixed files (HR packs, claims, accounting papers), you can describe the fields once instead of rereading every file. Each value comes back with a score and the source page, so people only review the weak extractions. The public templates AWS lists are mostly US tax forms and identity documents: for Belgian or French paperwork, you define the fields yourself. AWS states no price and no GDPR or EU AI Act terms here; files go through S3, so region, processor contract and processing records stay your responsibility.

What you can do

  • List which PDFs still mix several document types and which fields staff still retype by hand.
  • If you already use Amazon Textract, review the Insights API and the public templates AWS describes (passport, birth certificate, US tax forms).
  • Before sending personal documents, confirm the AWS region, your processor contract, and a human review rule based on the confidence score.
Intro call · 20 min · free

Let's talk about the time you could save.

In 20 minutes, we find where AI could save you time or money. Free, no commitment.

[email protected] Belgium · Europe (remote)
Book a call Call