Skip to main content

​
1. How AI interacts with you when unsure

In daily customer communication, AI will try to complete replies independently. But if it encounters uncertain situations, it will not answer rashly, but will ask you (Leader) for advice immediately. It will sync two messages to you:
  1. Question Forwarding Notice: The customer’s question is forwarded as is, ensuring you see the customer’s needs completely.
  2. Thinking Draft Suggestion: A draft reply generated by AI based on existing knowledge and context, asking you to judge whether it is appropriate.
Your handling method:
  • Quote “Question Forwarding” to answer directly → AI will forward your reply to the customer as is, ensuring accurate and consistent expression.
  • Quote “Thinking Draft” and reply “1” or any content → Regarded as agreement, AI will send the draft to the customer.
  • Manually switch back to AI account to reply → Suitable for complex, sensitive scenarios or those requiring more personalized expression.
The above mechanism not only avoids incorrect answers affecting the experience but also allows AI to continuously learn your decision logic in practice.

​
2. How to Quickly Close and Reopen

In different scenarios, you may need to temporarily close the AI or let it re-enter the working state.
  • Pause Immediately: #stop → AI remains silent to customers but will still respond to Leader’s messages.
  • Continue Working: #work → AI resumes external service and returns to work mode.
After logging in, AI is silent by default; only after you enter #work will it actively serve externally. This design is to avoid unauthorized disturbance. In addition, you can set silent white/blacklists for specific groups or contacts in the management backend to precisely control which objects can be responded to by AI.

​
3. How to Teach It More and Keep Getting Smarter

The core value of AI lies in continuous learning and gradually adapting to your business. Its main learning sources include:
  • Communication between colleagues and customers → Learning real business context and expression habits.
  • Direct guidance from Leader → Forming more robust judgment standards through your confirmation, negation, and rewriting.
  • Backend Knowledge Base Management → Systematic maintenance and updating to ensure knowledge traceability and governance.
  1. Have colleagues complete authentication (#auth → #peer) to avoid being misjudged as customers, ensuring AI correctly identifies internal knowledge and learns from it.
  2. Pull AI into more real business groups to absorb typical cases and accelerate coverage of common problem domains.
  3. Conduct more “quote-style” guidance with AI. Your rewriting and comments will accumulate as reusable learning materials.

​
Accurately Add New Knowledge

In addition to passive automatic learning, it is recommended to actively “feed knowledge” to build a more accurate and stable knowledge skeleton. Method 1: Backend Batch Management
  • Perform structured upload, hierarchical classification, and version management at admin.marsmind.cc / .co, suitable for large-scale knowledge governance.
Method 2: Single Entry (#chatknowledge)
  • Suitable for instant, precise supplementation of hot knowledge points.
  • Operation steps:
    1. Use Leader account to enter #chatknowledge (or #TrainingMode)
    2. Enter customer question
    3. Enter standard answer
    4. Wait for AI confirmation of inclusion to take effect
Please try to use real customer tone and supplement necessary context (object, limiting conditions, examples). This will significantly improve matching accuracy and transferability.

​
4. How It Responds in Groups (Group Chat Intelligent Response)

When AI is invited to join a group chat, it will follow the principle of “moderate participation”:
  • If a customer asks a question and no one responds for about 3 minutes, AI will try to supplement the answer;
  • If unsure whether it can answer accurately, AI will ask the Leader first to avoid misleading.
Function Switch:
  • Enable: #enable_hot_assistant
  • Disable: #disable_hot_assistant
Once enabled, AI will continuously track and analyze group information, which may increase package usage. Please control according to scenarios and budget.
The management backend supports batch setting of identities (Leader/Peer/Customer) for members, avoiding confirmation by individual instructions, and also reducing the probability of accidental triggering in groups.

​
5. Use of Weekly / Daily Reports

AI can generate work reports periodically to assist the team in synchronizing information and reviewing.
  • Daily Report On/Off: #enable_daily_report / #disable_daily_report
  • Weekly Report On/Off: #enable_weekly_report / #disable_weekly_report
Daily reports are suitable for high-frequency monitoring and troubleshooting; weekly reports are suitable for periodic review and trend tracking. You can enable flexibly according to management preferences.
For a complete list of commands and detailed instructions, please refer to MarsMind AI Interaction Methods.

​
7. Collaboration Best Practices and Support

Best Practices (For Reference)
  1. Familiarize yourself with basic commands and common operations first, so that you and your team gradually get used to the interaction method with AI, understanding when it needs your confirmation and when it can complete tasks independently.
  2. After becoming familiar, configure necessary identity relationships and permission boundaries combined with actual scenarios. Clarify who is Leader and who is Peer, and set groups or personnel that need to be silent in the backend in advance to avoid AI mis-sending to inappropriate objects.
  3. It is recommended to start pilot operation from a small number of core groups or customers for 1–2 weeks. Observe knowledge coverage, response effects, and colleagues’ adoption, adjust according to results, and then gradually expand to more groups and business scenarios.
  4. In the long run, the goal is for humans and AI to gradually form a stable collaborative relationship. You and the team provide instructions, feedback, and scenarios, while AI accumulates through execution and learning. By regularly optimizing and supplementing the knowledge base in the backend, make this collaboration smoother and truly become a reliable work partner.
Support and Contact We sincerely hope you can truly treat MarsMind AI as a colleague to collaborate, like him, guide him, and gain continuous value from him.
If you have any questions or suggestions during the collaboration process, please feel free to contact: