# Document processing isn't one problem -- it's three

DevFeed: [Document processing isn't one problem -- it's three](<https://devfeed.tech/articles/document-processing-isn-t-one-problem-it-s-three-57223.md>)

Original publisher: [Read original article](<https://www.thoughtworks.com/insights/blog/generative-ai/document-processing-is-not-one-problem-it-is-three>)

Author: Alfred Subietas I Oliveras

Published: 2026-02-26T00:00:00Z

Content type: article

Language: en

Sources: [Thoughtworks Insights](<https://devfeed.tech/sources/thoughtworks-insights.md>)

Topics: [Parsing](<https://devfeed.tech/topics/parsing.md>), [Parser](<https://devfeed.tech/topics/parser.md>), [pdf](<https://devfeed.tech/topics/pdf.md>), [processing](<https://devfeed.tech/topics/processing.md>), [layout](<https://devfeed.tech/topics/layout.md>), [structure](<https://devfeed.tech/topics/structure.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Microsoft](<https://devfeed.tech/topics/microsoft.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [auditability](<https://devfeed.tech/tags/auditability.md>), [batch](<https://devfeed.tech/tags/batch.md>), [blog](<https://devfeed.tech/tags/blog.md>), [document-processing](<https://devfeed.tech/tags/document-processing.md>), [generative-ai](<https://devfeed.tech/tags/generative-ai.md>), [layout](<https://devfeed.tech/tags/layout.md>), [nested](<https://devfeed.tech/tags/nested.md>), [risk](<https://devfeed.tech/tags/risk.md>), [text](<https://devfeed.tech/tags/text.md>)

## AI overview

This article explains why enterprise PDF document processing is not a single task. It examines failures in layout-aware text extraction, table structure preservation, diagram interpretation, and source traceability when using Copilot, and argues that these require distinct approaches.

## Source excerpt

Document processing isn't one problem -- it's three