Create an AI Assistant

1. Create an AI Assistant

Navigate to "AI Features" in the left menu and select "AI Assistant", then click "+Create AI Assistant" in the top right corner.

2. Name Your AI Assistant

Select the "Basic Settings" tab and enter a name in the "AI Assistant Name" field. You can name it based on its main task, such as XX AI Customer Service, XX Regulatory Query Helper, or XX Project Smart Assistant.

You can create multiple AI assistants under one account (limited by your subscription plan)

3. Choose RAG to Make Your AI Assistant Smarter and More Accurate

What is RAG

Think of RAG as a combination of a "conversational assistant + skilled research librarian".

A regular AI assistant is like someone with excellent memory who's good at storytelling but can only share previously learned knowledge. However, when an AI assistant is equipped with RAG technology, it's like having an assistant who first goes to the library to find the latest information before answering questions, then organizes the findings into clear responses for you.

In the MaiAgent platform, this "library" is our Knowledge Base. The AI assistant uses RAG technology to find relevant information from the knowledge base, making responses more accurate, timely, and relevant to needs.

The knowledge base setup will be explained in detail in the next section.

MaiAgent RAG not only includes the RAG technology mentioned at the OpenAI Developer Conference but also combines various classic NLP algorithms and proprietary retrieval technologies. Through internal datasets compared with OpenAI RAG's response accuracy, both achieve 95% response precision.

RAG Setup

Select the "RAG Settings" tab and choose different RAG (Retrieval-Augmented Generation) options from the "RAG" dropdown menu. If no special requirements, MaiAgent RAG is the default.

Compared to OpenAI RAG, MaiAgent RAG offers more additional features, flexible deployment options, handles more diverse data formats, and provides a more powerful retrieval and generation experience.

For detailed comparisons, see What is RAG? Comparison between Maiagent RAG and OpenAI RAG

4. Choose a Model - Select a Smart Brain for Your AI Assistant!

Purpose of Language Model Selection

Each AI assistant's performance largely depends on its brain - the language model (LLM) it uses. In this step, you can choose different types of models based on your needs, affecting response speed, comprehension ability, and depth and naturalness of answers.

Choosing the right model is like enabling high-performance mode for your AI assistant, creating the best experience tailored to your use case!

Key Factors in Large Language Models

Language Model Setup

Select the "RAG Settings" tab and choose different large language models from the "LLM Model" dropdown menu. If no special requirements, Claude 3.7 Sonnet is the default.

5. Create Role Instructions for Different Applications

To make AI assistants more suitable for different application needs, you can set "Role Instructions" to make their response style and content more appropriate for specific scenarios.

What are Role Instructions? Role Instruction Templates?

AI Tool for Generating Role Assistants

Hallucination-Free AI Response Mechanism

MaiAgent's "Hallucination-Free AI Response Mechanism" ensures high accuracy in AI responses. When facing uncertain or knowledge-exceeding questions, it honestly expresses its limitations rather than generating fictional answers, providing users with more reliable and trustworthy AI interactions. The importance for various industries and public sectors is explained below:

Industry Applications

Financial Industry Application:

When handling investment advice and risk assessment, AI must provide analysis based on actual data, avoiding misinformation that could lead to incorrect investment decisions. When information is insufficient or uncertain, the system will clearly indicate this to ensure investment reliability.

Healthcare Industry:

In assisting with medical diagnosis and medication consultation, AI systems must strictly follow established medical knowledge and not generate potentially misleading advice. For novel or unverified medical information, the system will clearly indicate the need for further professional consultation.

Manufacturing Industry:

In production process optimization and quality control applications, AI must provide recommendations based on actual production data and verified methods, avoiding production losses due to inaccurate predictions.

Education Industry:

In supporting teaching and answering student questions, AI needs to provide accurate knowledge rather than potentially misleading information. For complex or ambiguous concepts, the system acknowledges its understanding limitations.

Legal Industry:

When providing legal information and advice, AI must base responses on existing regulations and precedents rather than speculative advice that may carry legal risks. The system clearly indicates matters requiring further confirmation from professional lawyers.

Customer Service:

When handling customer inquiries, AI must provide accurate product information and service descriptions, immediately referring uncertain issues to relevant professionals to avoid misleading customers.

Public Sector Applications

Government Policy Consultation:

When providing policy information and service guidance to the public, AI must respond based on the latest and correct regulations and administrative procedures, avoiding outdated or incorrect information. When encountering complex issues requiring professional judgment, the system clearly recommends seeking assistance from relevant departments.

Public Service Decision-Making:

When assisting government evaluation of public construction and social welfare decisions, AI must analyze based on real data and research, clearly explaining uncertainties in predictions to ensure policy reliability.

Emergency Response Management:

When handling disasters, public health emergencies, etc., AI systems must provide accurate information and guidance, avoiding false information that could mislead the public and affect disaster response effectiveness.

Choose Appropriate Response Mode and Create Instructions

Select the "Response Mode" tab to meet various business needs, from free conversation to highly structured responses. Each mode has its characteristics and applicable scope, which you can flexibly choose based on actual usage scenarios.

Response Mode: General

Suitable Scenarios

Free response based on questions, AI generates the most appropriate response according to context and knowledge base content. Suitable for most Q&A scenarios.

Operation Process

Select the "Response Mode Settings" tab, choose "General (Default)" as the response mode, and enter your role instructions in the "Role Instructions" field. Output format can be plain text or JSON.

Application Scenario: Website Customer Service Assistant

If you want to create a website customer service assistant for "MaiAgent - AI Assistant Development Platform", you can input AI response settings in the "Role Instructions" field, clearly defining its response style and scope of responsibilities.

Response Mode: Template

When you choose "Response Mode: Template", your AI assistant's hallucination response rate will drop to zero, as it will strictly respond based on your established knowledge base and FAQ content, using classification principles rather than LLM-generated content. Using the template system guarantees 100% hallucination-free responses.

Suitable Scenarios

Scenarios requiring unified response formats, such as standard consultation processes, reports, etc.

Case Operation

Let's say you want to create a "Tainan City 1999" customer service AI assistant to respond to citizens' municipal inquiries in real-time.

1. Select the "Response Mode Settings" tab and choose "Template"

2. Fill in Response Template (Using template system for responses)

If you haven't created a knowledge base and FAQ yet, the default response template instructions are as follows.

It's recommended to first add a knowledge base or FAQ to use response templates, and the format must be Excel, CSV, json, or jsonl table-type files.

For this scenario, I have an AI assistant "Tainan City 1999" and added "Tainan 1999 FAQ" to the knowledge base.

The FAQ fields include questions, answers, departments, categories, and publication dates.

It's recommended to upload Excel, CSV, json, or jsonl table-type files, and fields must have titles and corresponding content.

Now return to edit the response template, click "Initialize Response Template" in the top right corner, and you'll see the system generate response examples based on the documents you just uploaded to the knowledge base.

Now you can edit and format the response text.

The [] {} parts are system commands, so just handle the text portions.

📍 Loop: Enter the corresponding document filename (specify which document to generate responses from)

I modified the following parts:

• "Opening" content

• "Question" content

• "Category" content

• Removed "Department"

• Closing content

• Line breaks and formatting

When we return to the AI assistant Q&A interface and ask related questions, the AI assistant will respond completely according to the template format and FAQ document content we just edited.

3. Fill in Unable to Respond Template

When the AI assistant determines there's no relevant data, it will respond according to the "Unable to Respond Template".

Finally, click the "Save" button.

Response Mode: Hybrid

Suitable Scenarios

Combines general responses with templates, using template format for some content and free responses for the rest. Suitable for scenarios requiring partially structured responses.

Operation Suggestions

The role instruction content might conflict with "Response Mode: Template", so when choosing "Response Mode: Hybrid", it's recommended to write role instructions focusing on general principles.

For example, the assistant's tasks, interaction principles, communication attitude, etc.

Response Mode: Workflow

Suitable Scenarios

Suitable for specific work task scenarios such as knowledge management, data summarization, and planning ideation.

Operation Process

Go to the "Response Mode Settings" tab and select "Workflow" as the response mode. This mode is suitable for task-oriented application scenarios such as:

• Knowledge Management: Help compile, retrieve, or maintain internal knowledge data

• Information Summary: Quickly organize document highlights or meeting minutes

• Planning & Writing: Help brainstorm ideas, draft proposals, etc.

Choose corresponding workflow modules based on actual business needs to maximize the AI assistant's effectiveness in specific tasks.

Application Scenario: Copywriting Assistant

Let's say you're a food company's marketing personnel writing promotional copy for a newly launched healthy snack.

First, you can select "Writing Assistant" in the "Workflow"

Then, you can submit your writing requirements in the AI assistant Q&A interface. The AI assistant will guide you through filling in key information needed for copywriting, such as:

• Copy topic

• Writing style

• Target audience

• Copy length

AI will generate copy options based on your settings that match the context and communication needs, helping you brainstorm quickly and improve writing efficiency.

Response Mode: Agent

Suitable Scenarios

In daily business operations, employees often need to organize data, conduct data analysis, and create reports and analysis reports in response to routine business needs or tasks assigned by supervisors. Common query questions include:

• "Which product had the highest sales last month?"

• "Please list all customers with sales over 100,000"

• "How's the revenue trend for the past three months?"

When these questions are handled by non-technical personnel, they often need to rely on the data team to write SQL queries, making the process time-consuming and inefficient.

Now, through MaiAgent's Agent mode, the system can automatically convert natural language questions into corresponding SQL syntax and query the database in real-time for quick analysis results.

This feature is particularly suitable for scenarios requiring real-time queries and data insights, such as report analysis, operational indicator tracking, and data queries. It allows non-technical users to easily access data, enabling more intuitive and efficient data-driven decision-making processes.

Operation Process

1. Go to the "Response Mode Settings" tab and select "Agent" as the response mode

1. Choose Database URL

MaiAgent

If you haven't created a database yet, you can directly upload Excel files, and the system will automatically convert them into a queryable database format

📍Note: Text-to-SQL Q&A function will not be available without uploaded data

Operation method: Upload Excel or CSV to the knowledge base, and the system will automatically create a database

• Table format considerations

  • Each sheet must contain only one table

  • The first row must be field names

  • To improve AI data retrieval accuracy and efficiency, it's recommended to annotate data formats in field names, e.g., price(int), date(datetime)

  • Special formats (e.g., currency, percentages) should be standardized before conversion

  • Ensure field names match actual content and use correct data types

📍Note: If you switch from an existing AI assistant response mode (e.g., from "General" to Text-to-SQL mode), please make sure to re-upload the Excel or CSV files you want to convert to a database. The system will create a queryable database based on the newly uploaded files to support SQL query functionality.

MySQL

Connect to an existing MySQL database by pasting the MySQL database connection string

📍Note: Ensure the database URL is accessible by MaiAgent services

PostgreSQL

Connect to an existing PostgreSQL database by pasting the PostgreSQL database connection string

📍Note: Ensure the database URL is accessible by MaiAgent services

Application Scenario: E-commerce Product Sales Data Query

Let's say you're an e-commerce platform's marketing personnel wanting to quickly query product sales data

First, in the "Response Mode Settings" tab, select "Agent" as the response mode

Then, you can choose to upload Excel files to the knowledge base, and the system will automatically convert them into a queryable database format

You can also directly ask your company's technical staff to provide MySQL or PostgreSQL connection strings.

Assuming we've obtained a PostgreSQL connection string, select PostgreSQL from the database URL dropdown menu, paste the connection string, and click save.

Once setup is complete, you can enter questions in the AI assistant Q&A interface, such as

"What are the top three items by website sales, excluding shipping fees?"

6. Pre-assign Assistant Permissions

RBAC-based Permission Management Architecture and Description

Select the "Permission Settings" tab to set which members need to be pre-assigned access to the assistant. By default, it's added to all roles, but settings can be modified based on usage scenarios.

7. Complete Your AI Assistant Creation

After entering all the above steps, click the blue "Confirm" button in the bottom right corner of the dialog to complete the setup 🎉.