AIDA is a cutting-edge AI assistant integrated into the AODocs Document Management System (DMS). Designed with a Retrieval-Augmented Generation (RAG) architecture, it combines advanced search capabilities with Large Language Models (LLMs) to provide precise, context-aware answers based on your library's content.

Automatically generated table of contents

Prerequisites

AIDA must be activated:

• on your tenant by AODocs support
• in your library by one of your AODocs super administrators 

When AIDA is activated, the Ask AIDA button appears in your library interface.

How AIDA works

AIDA doesn't simply guess answers based on its training. Instead, it follows a multi-layered process to ensure every response is grounded in your library's specific data.

In the example below, AIDA's answers will be based on documents in the S0 - Contract Management - Lataù library only. Only documents the user can access will appear in the responses. Learn more: Trust, security, and data scope.

Chain of thoughts

• What is it?
  Chain of thoughts (CoT) is a logical reasoning technique that breaks down complex questions into a sequence of smaller, manageable sub-problems.
• How it improves accuracy
  By "thinking aloud" through intermediate steps, AIDA avoids jumping to premature or incorrect conclusions. This is particularly effective for "multi-hop" questions where the answer must be synthesized from several different document sources.

Retrieval-augmented generation

• What is it? 
  Retrieval-augmented generation (RAG) is a hybrid architecture consisting of a retriever (which finds relevant documents) and a generator (the large language model).
• How it improves accuracy
  Standard AI models are frozen in time based on their latest training date. RAG allows AIDA to look up your live documents in real-time. This grounding in current data significantly reduces hallucinations (made-up facts) and ensures that if a document was updated five minutes ago, AIDA uses the latest version.

Note about up-to-date reliability: Every time you modify metadata, change an attached file, or perform a workflow transition, the document is automatically reindexed. This keeps the RAG retriever perfectly synced with your current library state.

The reranker

• What is it? 
  The reranker is a quality control layer that re-evaluates the documents found during the initial search.
• How it improves accuracy
  A standard search might find 50 documents that contain a specific keyword, but not all of them are relevant to your question. The reranker performs a deep semantic analysis to prioritize the most pertinent results, ensuring that only high-quality context is fed to the AI generator.

Hybrid search

AIDA uses a hybrid search to ensure no documents are missed. This means it user both semantic and full-text search. Learn more: Search and retrieval.

Main capabilities and use cases

AIDA uses a variety of tools to interact with your data.

Search and retrieval

• Full-text search
  Searches for literal keyword matches in the content of attached files and metadata. This is essential for precision tasks, such as finding a specific SKU, contract ID, or name.
• Semantic search
  Uses a vector database to find documents based on understand the meaning and intention behind the words, rather than just key words. For example, it can find "billing issues" even if your document only mentions "invoice discrepancies". The semantic search requires vector indexing to be activated.
• Natural language queries: Users can search using conversational phrases like "Find me the documents I created last month" or "Find the signed contracts from last year".

Content generation

There are three steps to generating content with AIDA:

• Ask a question: AIDA provides answers to complex questions by synthesizing data from both full-text and semantic searches. 
• Create a document: AIDA generates a brand new document based on the AI’s response.
• Attached file: AIDA appends the AI-generated text as a new attached file to an existing document.

Trust, security, and data scope

• Permission-based access: AIDA strictly respects AODocs security. A user can only ask questions about documents they already have permission to access.
• Data privacy: No training is conducted using your proprietary data.
• History retention: Past chats are stored and are accessible for 90 days from the date of creation.

Frequently Asked Questions (FAQ)

Do you train AIDA on our questions or data?

No. AIDA uses your trusted LLM to process queries. Your data is used as context for the RAG process to generate answers but is never used to train or improve the underlying AI models.

Which AI models does AIDA support?

AIDA supports the latest models from several providers, including:

• OpenAI: GPT 5.1, GPT 5.2
• Google: Gemini 2.5 Pro, 2.5 Flash
• Mistral: Mistral Medium 3

Note: During a free trial, AIDA defaults to an AODocs-validated GPT model.

What is the scope of AIDA’s search?

By default, the scope of a question is the entire library where you are working. However, the specific scope varies depending on where in the interface you are asking the question. For example, if use AIDA in a document, the answers will be based on the content of the attached files and the properties of the current document.

How can I get the most accurate answers?

Be as descriptive as possible in your prompts. For example, instead of typing "Pipeline ABC," ask "Explain how to design the pipeline ABC" to ensure AIDA understands if you want a search result or an explanation.