Just tell it what's in the folder.
Invoices, contracts, wiring diagrams, email, manuals: 15 folder types, and for each one the Box already knows what to look for and what to pull out. You point at the folder, the rest is automatic — and none of it leaves the building.
Create the folder
Upload documents or sync a cloud folder: Drive, OneDrive, SharePoint, Dropbox or your corporate NAS.
Pick the analysis type
Invoices, contracts, PLC, wiring diagrams… just pick the type: the Box already knows what to look for, with no settings to study.
Ask questions
The folder answers back: ask in plain words what it holds, what is coming due and where it says so — with the document it came from.
Administrative & legal documents
Invoices, contracts, HR files and email archives: metadata extracted automatically and relationships mapped between entities.
Invoices & receipts
Table and amount extraction with visual analysis of scanned documents. Every invoice becomes a structured, searchable record.
What it pulls out: Date, number, amount, supplier, VAT number
Contracts & legal
Knowledge graph of parties and clauses: the system links counterparties, signing dates, expirations and values across the whole archive.
What it pulls out: Parties, signing date, expiration, value
Human resources
CVs, employment contracts and reviews with table extraction and structured fields ready for filtering and search.
What it pulls out: Name, role, hire date, department
Email archives
Threads and attachments indexed with a conversation graph: who wrote to whom, when, and about what.
What it pulls out: Sender, recipient, date, subject, attachments
Structured data & catalogs
When fields and tables matter: turbo pipeline for high volumes and custom metadata schemas.
Custom metadata extraction
Define your own field schema and the system extracts them from every document with the fast pipeline, ideal for large volumes.
Tabular data
Table-heavy documents: price lists, reports, BOMs. Tables are reconstructed with queryable rows and columns.
Product catalogs
Catalogs with datasheets, images and price lists: table and visual extraction combined, with structured product fields.
What it pulls out: Name, code, price, category, availability
Project documents
Specifications, minutes and deliverables linked in a graph: projects, owners, statuses and deadlines always traceable.
What it pulls out: Project, owner, status, deadline
Research & papers
Scientific papers and studies with relationship extraction between concepts, authors and citations for truly semantic search.
Technical & industrial
From the plant floor to the engineering office: PLC programs, electrical schematics, system logs and machine manuals.
PLC / SCADA programs
Upload Rockwell Studio 5000 projects (.ACD/.L5X) and Ignition tag exports: the AI diagnoses alarms explaining cause, reset conditions and consequences, rung by rung.
Electrical schematics
Component and connection recognition on schematics: ask where a signal runs or which terminals a relay touches.
System logs
Log files classified by error signatures and anomalies: troubleshooting starts from patterns, not manual grep.
Technical manuals
Machine manuals and operating instructions with automatic description of figures and diagrams, so even images become searchable.
General content
For everything else: mixed archives, images and heterogeneous documents with hybrid semantic + keyword indexing.
Images & media
Photos, renders and scans analyzed with vision models: search by content, not by filename.
General documents
The universal preset: hybrid indexing (semantic + keyword) for mixed archives of PDFs, Office files, text and web pages.
It even reads PLC logic.
Load a Rockwell Studio 5000 project and ask about an alarm: the Box tells you which condition trips it and what it takes to clear, citing the exact point in the program.
Discover AI for industry❯ Why did alarm 4.16 trip?
Alarm: ALL_MEM[4].16 — Press 2 — hydraulic consent missing
Cause: TON timer T_CONSENSO (5 s) expired without the minimum pressure signal PS_02.
Reset: Pressure restored above threshold + rearm button on the panel.
Consequences: Press 2 automatic cycle stops; the downstream line goes into standby.
❯ Which terminals does relay K12 touch?
K12 drives pump M3 through terminals X1:4 and X1:5 — sheet 12, column C. The coil is on sheet 12, driven by S1 under fuse F1.
The wiring diagram reads itself.
Components, tags and connections recognised straight off the PDFs: ask where a cable runs or what a contactor drives and get the answer with the sheet it came from.
- Visual recognition of components and connections on every sheet
- Answers referencing sheet, column and component designation
- Schematics, machine manuals and PLC programs in the same knowledge base
What happens to every document.
Every file goes through the Box and comes out readable, searchable and citable. The folder type decides which steps are actually needed.
Conversion & parsing
PDF, Word, Excel, PowerPoint, .msg emails, .p7m digital signatures, HTML and text are normalized into pages and blocks.
Smart OCR
Text-less scans are recognized with local vision models, on dedicated GPU.
Tables & figures
Tables are rebuilt row by row; figures and diagrams get described so they become searchable.
Structured metadata
Entities and custom fields extracted from every document: dates, amounts, counterparties, product codes.
Knowledge graph
Entity relationships are linked in a graph: cross-archive questions, not just single-file lookups.
Hybrid index
Dense semantic embeddings + sparse keyword vectors: search understands meaning without losing exact terms.
Specialist analysis
PLC programs, schematics and logs go through dedicated analyzers with on-demand diagnosis right in chat.
Everything on-premise
Every stage runs on the Box, inside your network: no document ever leaves the company.
Not sure which folder to start with?
Tell us how your archive is put together and we'll tell you what the Box can pull out of it, and where it's worth starting.