Loading ...
AI isn't a separate screen; it lives inside the field where you enter data.
In Documenos®, AI is not a separate window you open to ask questions. It is bound to the definition of the field where you create content: the prompt knows which fields to read from and which fields to write the result to.
In practice this means users don't copy text into another application and paste the result back. The job is finished inside the form they are already filling in.
Six providers are supported: OpenAI, Azure AI, Google Gemini, Google Vertex, HuggingFace and Ollama. With Ollama the model runs on your own server, so document content never leaves your organisation for the AI service.
You can define an unlimited number of connections, and each one chooses its own model. So you can, for example, use a local model for sensitive content and a cloud provider for everything else.
The AI prompt doesn't belong to a single screen; the same structure is used wherever content is created:
The document's header fields.
Dynamic fields and tables in form templates.
The template's action fields, at any step of the flow.
The same structure applies to these field definitions too.
Input and output fields can be multiple: the prompt takes information from several fields on the screen all at once, sends it to the AI agent, and can write the response back into multiple fields separately.
So a prompt built as read these fields, fill in those fields becomes a single click for the user.
The same approach works for documents written as free text: data in the document's header fields becomes input to the prompt, and the text is generated from that information.
The organisation defines the template once, so users don't have to rewrite the same prompt every time.
AI doesn't just write into fields; it also makes the content of files added to the archive searchable.
Speech in audio and video files is transcribed in 100+ languages and added to the search index, so you can find a phrase spoken in a meeting recording by searching for it.
Text is extracted from scanned documents and photos, so even a paper stored only as an image shows up in searches.
Human faces in photos are detected and similar faces are suggested. The user confirms each match; the system never automatically links a person it sees for the first time to an existing person. You control the similarity threshold.
Documenos® hosts an MCP (Model Context Protocol) server. Users generate their own access key, and the AI assistant they use can connect to Documenos® to list documents, tasks, processes and workflows and act on them.
The assistant works with the user's own permissions: the AI assistant cannot access any document or information the user can't. Its actions go through the same permission checks and are logged.
