Assistant Setup Guide

What is an Assistant?

A DataQI assistant is a tailored AI interface trained to understand and respond using your internal content. Assistants can be built to answer questions, create documents, or help retrieve information from connected files.

When to Create a New Assistant

You have a specific audience (e.g. HR, Legal, Ops)
You want to use different data sources or templates
You need a different tone, language style, or purpose
You want to limit access to a defined group of users

Setup Steps

1. Choose Assistant Type

Select from:

Chat - for natural language Q&A
Document Writer - to generate structured outputs
Search - for searching and retrieving documents

Assistant Type Selection

2. Name Your Assistant

Give your assistant a clear, descriptive name. This helps users know what it’s for (e.g. “HR Policy Chat”, “Sales Proposal Writer”).

3. Model Selection

If your plan includes access to multiple AI models, you’ll see an option to choose a model during assistant creation. Select the model that best fits your task:

Use lighter, faster models for quick summarisation or general responses
Choose higher-accuracy models for legal, policy, or high-stakes content

Your admin team can advise on model availability and usage based on your licence tier.

4. Use Existing Data Sources

Before creating your assistant, ensure your content has been connected to DataQI:

Upload documents (PDF, Word, TXT, Markdown)
Link folders from SharePoint or Google Drive
Add public content via web scraping
Connect SQL databases (Microsoft SQL, PostgreSQL, MySQL) for real-time data querying

Once data is connected, you can assign it to an assistant during setup.

5. Configure Assistant Settings

Use the unified Configuration page to customize your assistant’s behavior and capabilities:

Assistant Name - Clear, descriptive name for easy identification
Description - Define the assistant’s role and purpose to improve context and response accuracy
Instructions (Optional) - Add specific rules or behaviors (e.g., “Always cite sources when providing information”)
Tone of Voice (Optional) - Choose the communication style (Professional, Friendly, Technical, or Concise)
AI Model - Select the underlying model powering your assistant
Configure Knowledge Mode - Choose whether to use only your private data or include global knowledge
Enable Hallucination Guardrail - Add safety measures to flag potentially inaccurate responses
Set up Intent Detection - Allow the assistant to automatically detect when to query live data sources

6. Upload Templates (if applicable)

If you’re creating a Document Writer, upload your .doc, .docx, .txt, or .pdf templates. These will define the output format.

7. Test Your Configuration

Before sharing your assistant, test its capabilities:

Ask sample questions to verify it uses the right data sources
Test SQL connections with data-related queries (if applicable)
Verify Knowledge Mode behavior matches your expectations
Check Hallucination Guardrail warnings appear when appropriate

Use the Privacy tab to share with users or groups:

View access - to use the assistant
Edit access - to change its configuration or content

Best Practices

Create focused assistants (e.g. one per department or process)
Use consistent naming conventions
Keep prompts short and specific
Configure appropriate Knowledge Mode for your use case
Enable safety features like Hallucination Guardrail for high-stakes content
Test thoroughly before sharing with users
Review feedback regularly to improve responses

Advanced Features

For more detailed information on the new features:

SQL Connections - Connect to databases with automatic intent detection
Knowledge Mode - Control data source usage
Hallucination Guardrail - Safety and accuracy features

Who Can Create Assistants

Only users with the relevant permissions (as assigned by your admin team) can create or configure new assistants. If you’re unsure, contact your organisation’s platform administrator.