← Articles

Using AI-Assisted Replies for Business Hours & Location Queries in Local Services

Learn how local service businesses can use GoodChat's knowledge base and AI-assisted reply mode to handle frequent inquiries about business hours and locations accurately, while maintaining human oversight.

Example Scenario: Customers Frequently Ask About Store Opening Hours

In the local life services industry, such as restaurants, beauty salons, or gyms, online customer acquisition channels often face a high volume of repetitive inquiries. For example, a potential customer enters a conversation via a website floating window and directly asks, "What time do you close today?" or "How do I navigate to your store?" These types of questions usually do not involve complex business procedures, but if left unattended during non-working hours, they can easily lead to lost leads.

At this point, relying solely on manual customer service may result in delayed responses during peak times; relying entirely on traditional bots may provide incorrect answers due to outdated information (such as adjusted hours after holidays). Therefore, adopting an "AI-assisted + human backup" strategy is more reliable.

Prepare an Accurate Static Knowledge Base

To effectively cover the above scenario, operators must first organize basic factual data. In the "AI Customer Service" module of GoodChat, it is crucial to maintain dedicated knowledge base entries.

  1. Clear Topic Segmentation: It is recommended to split topics like "Business Hours," "Specific Address and Directions," and "Parking Policies" into independent knowledge points. This ensures that when users ask for directions, the AI does not confuse the answer with parking fee issues.
  2. Text Content Verification: The body text of the knowledge base must be verified latest text provided by the operator. For example, "Monday to Sunday 10:00-22:00, holidays subject to separate notice." Note that these descriptions are used to help the AI select appropriate knowledge bases, not to let the AI speculate. If a store temporarily closes, the status of the entry must be updated or the text modified in the backend promptly.
  3. Publishing and Selection: After editing, you must click "Publish Knowledge" and then check/select the corresponding knowledge base in the bot configuration. Only after completing this step can the AI retrieve this information in subsequent conversations.

Configure Assisted Reply Mode to Balance Efficiency and Safety

For questions involving specific in-store experiences, although AI can provide standard answers, it is recommended to enable "Assisted Reply" mode to avoid misleading customers and causing complaints.

In this mode, when visitors raise questions about time or location, the system generates suggested replies based on published knowledge bases. However, this reply is not sent directly to the visitor but appears in the agent's workbench, awaiting human confirmation. Customer service agents only need to quickly review it; if correct, they send it with one click; if there are special circumstances (e.g., early closure due to equipment maintenance today), they can manually correct it before sending.

This mechanism leverages AI's fast retrieval capabilities to reduce the burden of manual typing while retaining human control over the final output content, aligning with the product principle that "reception handles inquiries and communication, while business outcomes are confirmed manually by merchants."

Understand Billing Logic and Applicable Boundaries

Many merchants worry that using AI will generate uncontrollable high costs. In fact, GoodChat's basic online customer service system is free and includes 2 agent seats. AI features are billed based on actual usage.

This means that every model call—whether extracting answers from the knowledge base or performing semantic matching—incurs corresponding input and output Token consumption fees. Even in "Assisted Reply" mode, triggering model inference to generate draft suggestions counts towards usage. Images and other multimedia content are also billed according to the actual usage reported by the supplier, rather than simply per image.

It is important to note that this solution applies to standardized information queries. If customers ask about personalized reservation success/failure, specific bill details, or refund progress, these belong to dynamic business data and are outside the scope of static knowledge bases. In such cases, if the AI cannot find exact evidence from existing materials, it should guide the transfer to manual customer service for further verification. Guessing policies or promising automatic completion of business operations is strictly prohibited.