Document Processing

Employees use AI to read, summarize, classify, and extract information from internal documents and contracts.

Business impact

• Daily active usage — LVMH reported roughly 4,000 to 5,000 daily GPT-4 users.
• Retrieval accuracy — Mitsubishi Electric reported initial RAG accuracy of 70% to 80%.

Data requirements

• Internal documents (text) — The workflows use internal documents, contracts, manuals, and design files.
• PDF documents (text) — One deployment explicitly mentions PDF documents.

AI methods and techniques

• Information extraction — Used to pull needed details from documents for review or downstream tasks.
• Document summarization — Used to condense contracts and other documents into shorter forms.
• Retrieval-augmented generation — Used to retrieve relevant design documents and answer engineering questions from internal content.
• Natural language processing — Used in the document-processing workflow to analyze unstructured text.

AI models and model families

GPT-4

Industries

Real-world evidence

11 documented case studies on record.

Companies using this: JPMorgan Chase, LVMH, Lockheed Martin, Mitsubishi Electric, Schneider Electric, Philips, MetLife, Assessor-County Clerk-Recorder (ACR) Office at a large Southern California County, Serious Fraud Office, Shell, CRIF.