OCR & AI

Document AI glossary

The useful terms for talking about AI applied to business documents — without unnecessary jargon.

By DocPilot team2 min read
Illustration of a glossary of document AI terms

This glossary complements the AI for business documents guide and the FAQ. Where useful, each entry points to a practical use.

A — C

  • **Automation (document)** — chaining receipt, extraction, routing and notifications without manual re-keying (the DocPilot process).
  • **Classification** — proposing a document type (contract, quote…) (article).
  • **Citation / source** — the passage of the document used to justify an AI answer (questions).

D — G

  • **DocpilotAI** — DocPilot’s assistance layer: document typing, extraction, questions on selected documents.
  • **Extraction** — automatically filling in structured fields (dates, amounts, parties) (article).
  • **DMS (document management system)** — known in French as GED; a single system of record with access rights and search (enterprise DMS).
  • **Governance (AI)** — rules that keep humans accountable for decisions (AI proposes / human approves).

H — O

  • **Human-in-the-loop** — AI proposes, a human confirms before any binding action.
  • **Metadata** — record fields that make the document filterable (type, party, status…).
  • **OCR** — optical character recognition: turning a scan or photo into text (how it works).

R — S

  • **RAG** — retrieval-augmented generation, drawing on passages retrieved from your document collection.
  • **Semantic search** — finding by closeness of meaning, not just by exact word (article).
  • **Summary (AI)** — an overview to prioritise reading; ≠ a legal decision (limits).
  • **Score (confidence)** — an indicator for targeting review (article).

W

  • **Workflow / approval process** — the sequence of human approvals after the AI proposal (approval).

Frequently asked questions

Which term should I start with?
DMS (system of record), OCR (reading), extraction (structuring), human-in-the-loop (approving).
What is RAG?
Retrieval-Augmented Generation: answering a question based on passages retrieved from your documents, rather than on the model’s memory alone.
Where can I learn more?
The AI for business documents pillar guide and the cluster FAQ.

From words to practice with DocPilot

OCR, DocpilotAI extraction, questions and workflows — on your first document flow.

Free resource

Download the checklist: 25 points to set up effective contract management.

Download the checklist