Documents
PDF, Word, plain text and Markdown.
Formats
- DOCX
- TXT
- MD
Document workspace
Work on your documents with any model.
Upload once and query across everything. Only the relevant snippets ever reach a model, so your full documents stay private.
How retrieval works
"Query your documents" is doing a lot of work in most product pages. Here is what actually happens between the upload and the answer.
01
The file is read, not stored blindly
Text is extracted page by page — slides slide by slide, spreadsheets as structured tables — and split into passages small enough to be precise.
Upload once
02
Each passage gets an index entry
Passages are indexed for meaning, not just keywords, so a question phrased differently from the document still finds it.
Semantic index
03
Your question retrieves a handful of passages
The index returns the few passages that actually bear on the question, from across every file you have uploaded.
Across all files
04
Only those passages reach the model
The model sees the question and the retrieved snippets — never the whole library. Change model mid-thread and the same rule holds.
Snippets, not libraries
Why it matters
Retrieval is the privacy boundary. Because only the matched snippets are sent, the volume of your material exposed to any model stays small and traceable, whichever model you pick.
What you can bring
Four kinds of material, one search. Everything below is read, indexed and quotable the moment it lands.
Data
Excel and CSV, read as structured tables.
Formats
- XLSX
- CSV
- JSON
- XML
Decks
PowerPoint, slide by slide.
Formats
- PPTX
Images
JPG and PNG, read by vision-enabled models.
Formats
- JPG
- PNG
Where the documents actually sit
Retrieval explains what is sent to a model. It says nothing about where the originals live, and that is usually the question a security review opens with.
In the deployment you chose
Your own cloud account, your own data centre with an air-gapped option, or a dedicated tenant we operate. The storage location is a property of the deployment, not a setting inside the product.
Behind the access rules you already have
Files belong to a workspace, not to the platform. Who can retrieve from them follows your role assignments, and every access is recorded against a named account.
Not in anyone's training data
Documents and the passages retrieved from them are not used to train models — ours or a provider's. We orchestrate models; we do not train them.
Removable
Deleting a file removes its passages from the index as well, so it stops being retrievable rather than merely disappearing from a list.
What retrieval will not do
It finds the passages that bear on a question; it does not read a document end to end. Ask for an exhaustive summary of a 300-page report and you are asking for a different operation, one that costs more and belongs in a different workflow. It can also miss: a passage worded very differently from your question may not surface, which is why the snippets it used are shown rather than hidden. And it can only answer from what has been uploaded — an answer that needs the contract nobody added will be confidently about the ones that were. Treat it as a fast, cited reader of your material, not as a system of record.
Questions about documents
What people ask before they upload the first file.
What file types does AI+ Pro support?
AI+ Pro supports a wide range of common document and text formats. You can upload and work with PDFs, Word documents, PowerPoint slides, Excel sheets, and many other text-based files such as CSV, TXT, Markdown, JSON, and XML. Image formats like JPG and PNG are also supported for vision-enabled models.
What is RAG and why does it matter?
RAG (Retrieval-Augmented Generation) lets AI pull information from your own documents before answering. It gives more accurate, context-aware results using your internal knowledge—not generic internet data. Only small relevant snippets are used, so your full documents stay private.
Can AI+ Pro search the web and my files at the same time?
Yes — AI+ Pro uses a Hybrid Answer Engine that combines web search with your internal documents in one response. You get answers that are accurate, up-to-date, and fully grounded in your organization's own information.
Where is my data stored?
You choose. AI+ Pro supports institution-owned storage (data stays on infrastructure you control), private cloud options (dedicated environments not shared with other organizations), and standard cloud storage (enterprise-grade encrypted storage managed by us). All options include end-to-end encryption and strict access controls.
Is AI+ Pro compliant with data protection regulations?
Our infrastructure is designed to support compliance with GDPR (Europe), CCPA (California), PDPA (Southeast Asia), and other regional data protection frameworks. However, compliance is a shared responsibility—we provide the technical controls and documentation, and you configure the platform according to your specific regulatory requirements.
Upload one file and see.
A single contract or report is enough to tell whether this reads your material properly. Start there.