AI GOUser Guide

AI Chat and Intelligent Assistant Operations

AI Chat (Cowork) is the core way of interacting with AI GO. Daily operational colleagues do not need to log in to the global console; instead, a dedicated AI collaboration chat window is embedded in your Custom App. You can collaborate with the AI assistant in real-time using natural language to complete daily tasks.

Entering the Chat Interface

Depending on your role, the way to enter the chat is as follows:

  • Management and Development Colleagues: Click "Cowork" under the "AI Agent" section in the left navigation bar.
  • Daily Operational Colleagues: After logging in to your dedicated Custom App, the AI collaboration chat window will be displayed directly in the sidebar or at the bottom right of the screen.

The chat interface contains the following core areas mentioned in the guided tour:

  1. Agent Connection Status: The status bar at the top displays the AI Agent's operation in real time. When the indicator light is green, the Agent is online and ready for your commands; you can view the name of the task the Agent is currently processing and the number of registered tools. If the Agent is offline, the status bar displays a red prompt.
  2. Conversation History: The main area in the center is the complete conversation history between you and the AI Agent. The Agent will report its thinking process, task planning, and execution results here, and explain error reasons and attempts at automatic recovery.
  3. Command Input: The input box at the bottom is where you type your questions or commands. Click the send button to submit a command, or click the stop button to abort a running task.
  4. Real-time Monitoring Panel: The collapsible panel at the bottom provides two real-time monitoring views:
    • Terminal: Displays the actual system CLI command output executed by the Agent, allowing you to transparently grasp every operational detail.
    • Screenshot: When the Agent operates a browser or graphical interface, it automatically captures the screen for your verification.

Prompt Techniques for Collaborating with AI

Providing structured and clear "Prompts" allows the AI to understand your needs faster and more accurately.

Suggested structure for high-efficiency prompts:

  • Specify Role and Goal: Tell the AI its identity and task (e.g., "As a financial assistant, help me calculate...").
  • Provide Specific Range and Conditions: Clearly specify time ranges, client names, product categories, or statuses (e.g., "Q2 this year", "Client A", "undelivered orders").
  • Specify Output Format: You can request the AI to reply in a table, list, or brief summary.

Question Templates

You can directly copy and modify the keywords in these templates:

1. Sales and Customer Data Queries (e.g., in Sales App)

  • Template 1: "Please list all consumption amounts for Client [Company Name] in the current year, and organize the date and amount of each order in a table."
  • Template 2: "Who was our salesperson with the highest performance last month? What was their main product sold?"
  • Template 3: "Query all quotation orders currently in 'Pending' status, and display the responsible salesperson and estimated amount."

2. Stock and Purchasing Queries (e.g., in Purchase App)

  • Template 1: "Check the current real-time stock quantity of raw material [Material Name]. Is it below the safety stock threshold?"
  • Template 2: "List all purchase orders and delivery statuses with Supplier [Supplier Name] over the past three months."

3. Financial and Analysis Decisions (e.g., in Accounting App)

  • Template 1: "Compare the revenue of Q1 last year with Q1 this year, analyze the growth rate, and provide a brief trend analysis."
  • Template 2: "List all accounts receivable that are currently overdue but unpaid, sorted from most overdue days to least."

Delegating Operation Tasks to AI

In addition to querying data, the AI assistant can help you perform simple daily operations and automatically write changes to the underlying Data Center. When you issue such commands, the AI will generate a draft or preview for your confirmation, and will only change the data after you click "Confirm Execution".

Operation command templates:

  • "Help me create a sales quotation for Client A, with 10 units of Product X at a unit price of 500 USD."
  • "Change the delivery status of Order 102453 to 'Shipped' and automatically send a notification email to the client."

File Center and AI Knowledge Base (RAG) Configuration

In addition to directly asking questions, administrators can upload and configure internal company documents via the "File Center", allowing the AI assistant to learn specific knowledge for Retrieval-Augmented Generation (RAG) smart answers.

1. Entering the File Center

  • Click "File Center" under the "Data Center" section in the left navigation bar.
  • The File Center provides a cloud drive-like interface, supporting folder creation, renaming, deleting, and batch file uploads.

2. "Set as Knowledge" Operational Steps

  • Upload your corporate documents (e.g., company policies, product manuals, FAQ summaries).
  • Right-click on the file card to open the context menu, and click "Set as Knowledge".
  • The system automatically uploads the file to the vector store of the AI service. Once processed, a green brain icon appears on the file, indicating the AI assistant has successfully loaded the knowledge.
  • From then on, whether in the global Cowork or in a Custom App, the AI assistant will automatically retrieve the document content to provide more accurate internal answers.

3. "Remove Knowledge" Operational Steps

  • If a document is outdated or no longer needed as an AI retrieval source, right-click on the file card and click "Remove Knowledge".
  • The file is removed from the AI assistant's search index, though the original file remains in your File Center.

4. Supported Formats and Limitations

  • Supported Formats: Includes .pdf, .docx, .xlsx, .csv, .txt, .json, .md, .zip, and code files.
  • Unsupported Formats: Media and vector image formats (e.g., .mp4, .mp3, .wav, .svg) cannot be set as knowledge.
  • Folder Limitation: The system does not support setting an entire "folder" as knowledge. You must enter the folder and set individual files.

FAQ & Troubleshooting

What should I do if the AI says it cannot find the data?

  1. Check Names and Spelling: Ensure the client name, product model, or order number you entered is exactly the same as registered in the system.
  2. Check Time Range: AI usually needs a clear time range to extract data accurately; try adding "Q1 this year" or "this week".
  3. Check Your Access Permissions: Your Custom App is only authorized to read specific tables. If you ask about unrelated payroll or inventory details in the Accounting App, the AI will refuse based on security policies.