#Customize HeiChat for Better Customer Service

#Purpose of this guide

HeiChat's answer quality depends on more than the AI model. It also depends on whether your store provides complete, accurate, and up-to-date business information. To make HeiChat behave more like a customer-service teammate who understands your brand, configure the following areas together:

• Use Basic Settings to define your brand identity and store context.
• Use AI Response to set the assistant's tone, language, answer length, and behavioral boundaries.
• Use AI Model to balance answer quality, speed, and token cost.
• Use Learned Products to make sure product information is available to the AI.
• Use Knowledge Base for shipping, returns, support procedures, and other information beyond product data.
• Use Gap Detection and Knowledge Gaps to find unanswered questions and add the missing knowledge.
• Use Live Support or Smart Tickets to provide a reliable human escalation path for complex issues.

The most effective approach is a continuous cycle: tell HeiChat who your store is and what it sells, define how it should communicate, then use real customer conversations to improve the knowledge it is missing.

---

#1. Start with Basic Settings: establish the service context

#Path

Shopify Admin > Apps > HeiChat > Setup > Basic Settings

#What to configure

In Core Setup Required, complete at least the following items:

1. Customize Your Bot

   • Set the bot name customers see.
   • Upload an appropriate bot avatar.
   • Select Set AI Style to continue configuring reply behavior in Advanced Settings > AI Response.

2. Store Name

   • Enter your official brand name rather than relying only on the Shopify domain prefix.
   • This helps HeiChat refer to your store naturally and consistently.

3. Store Introduction

   • Briefly explain what you sell, who you serve, your main product categories, and your usual price range.
   • Include shipping regions, handling times, important return rules, and key brand or service characteristics.
   • Use short paragraphs or bullet points instead of marketing slogans alone.

4. Customer Service Supervisor's Email

   • Enter a mailbox that your team actually monitors.
   • HeiChat can provide a dependable next contact point when a customer needs human help.

5. Product Learning

   • Confirm that product learning has completed or is progressing normally.
   • Do not judge product answers or recommendations before learning is finished, especially after first installation or a large catalog update.

6. Activate Chat Widget

   • Verify that HeiChat is activated in the storefront theme.
   • Customers cannot use the assistant if the widget is not active, even when the rest of the configuration is complete.

#Why this matters

Basic Settings provide HeiChat's first layer of context. Store Name and Store Introduction tell the AI which brand it represents and what it needs to know about the business. A monitored supervisor email provides a safe fallback for unresolved issues. Product Learning gives HeiChat access to the real catalog information it needs for product questions and recommendations.

Write the Store Introduction as information a service agent can use directly: what you sell, where you ship, how long fulfillment usually takes, and which items are not eligible for return. Generic statements such as “we offer high-quality products” do not help the AI answer customer questions.

---

#2. Configure AI Response: keep replies consistent and safe

#Path

Shopify Admin > Apps > HeiChat > Setup > Advanced Settings > AI Response

You can also reach this page from Setup > Basic Settings > Customize Your Bot > Set AI Style.

#Recommended settings

1. AI Output Mode

   • Streaming is recommended for most stores because the reply appears progressively and reduces the customer's perceived wait.
   • Use Complete Response only if your brand prefers to show a fully composed answer at once.

2. Reply Style

   • Select Serious & Professional for formal, regulated, or high-consideration brands.
   • A friendly, lighter style works well for many consumer brands.
   • Use high levels of humor carefully, particularly for refunds, complaints, and order issues.

3. Emoji Usage

   • Formal or premium brands may prefer Prohibited or Minimal.
   • Consumer brands can use Moderate, but frequent emojis in every reply can make support answers harder to scan.

4. Response Length

   • Start with Standard for a balanced default.
   • Use Concise for fast FAQs and mobile-first conversations.
   • Use Detailed for technical products, installation guidance, or complex policy explanations.

5. Reply Language

   • Choose Automatic for a multilingual audience.
   • Choose Fixed Language when your store must respond in one specific language.

6. Custom Instructions

   • Define a consistent brand voice, for example: be calm, helpful, and easy to understand.
   • State what HeiChat must not promise, such as unconfirmed delivery dates, stock, or policy exceptions.
   • Explain how it should handle uncertainty, for example: ask one clarifying question instead of making an assumption.
   • Specify that sensitive account, payment, refund-exception, legal, or medical issues require human support.

#Why this matters

Knowledge Base determines what HeiChat knows; AI Response determines how it communicates that knowledge. The same return policy can be presented in a structured, formal way for a premium brand or in warmer language for a lifestyle brand, but the policy itself must remain accurate.

Keep knowledge-gap criteria out of Custom Instructions. Rules for recognizing missing knowledge belong in Gap Detection, while AI Response should contain only customer-facing behavior and communication guidance.

---

#3. Choose an AI Model: balance quality, speed, and cost

#Path

Shopify Admin > Apps > HeiChat > Setup > AI Model

#What to consider

The AI Model page lets you test answers and choose the model that powers HeiChat. Models in the app are typically labeled by provider, tier, and token multiplier. When making a choice, consider:

• how complex your customers' questions are;
• whether you need stronger multilingual support or reasoning;
• how often the AI handles policies, installation instructions, or compatibility questions;
• the expected response speed;
• the token use per conversation and your total budget.

As a general approach:

• Basic models are well suited to FAQs, simple product information, and straightforward pre-sales questions at lower cost.
• Pro models can offer a practical balance between cost and answer quality.
• Advanced models are a better fit for complex policies, detailed support questions, and stronger reasoning needs.
• Premium models can suit high-value consultations or demanding technical support, but usually consume more tokens.

The available models and token multipliers can change as providers and HeiChat update their offerings. Use the current labels in the AI Model page as the source of truth, and test with your own store scenarios rather than choosing based only on a model name.

#Recommended testing workflow

1. Prepare real questions covering products, shipping, returns, promotions, order issues, and multiple languages.
2. Test a Basic or Pro model to establish the baseline.
3. Test an Advanced or Premium model against your more difficult questions.
4. Compare answer accuracy, completeness, tone, speed, and token consumption.
5. If several models give the same incorrect answer, review Knowledge Base or Learned Products before upgrading again.
6. Choose the model that meets your service standard at an acceptable cost, then enable it for the storefront.

#Why this matters

A more expensive model cannot reliably answer information that the store has not provided. If the AI lacks the return policy or a key product attribute, even a stronger model may still respond vaguely. First supply accurate business knowledge, then use model selection to improve the understanding and presentation of more complex questions.

---

#4. Confirm Learned Products: make product answers trustworthy

#Path

Shopify Admin > Apps > HeiChat > Setup > Learned Products

#What to check

• Whether product learning is complete.
• Whether important products are included within the learning scope.
• Whether product titles, descriptions, variants, materials, dimensions, and compatibility information are complete.
• Whether automatic synchronization has completed after product updates.

HeiChat automatically reads product changes from Shopify and synchronizes the updated product information. For small, routine edits, you normally do not need to trigger an action every time. If the store makes a large batch of product updates, adds many products at once, or changes important attributes such as descriptions, variants, materials, dimensions, or compatibility, manually select Refresh on the Learned Products page to make sure the latest information becomes available to AI replies and product recommendations in a timely manner.

#Why this matters

Product questions are different from general FAQs. Customers may ask which size suits them, whether an item is compatible with a device, which colors are available, or which product meets a particular need. These answers rely on current Shopify product data and should not all be maintained as manual FAQs.

If a product record is missing an important attribute, improve the Shopify product information first, then wait for automatic synchronization or manually select Refresh. Avoid repeatedly adding duplicate product details to Knowledge Base.

---

#5. Build a dependable Knowledge Base

#Path

Shopify Admin > Apps > HeiChat > Setup > Knowledge > Knowledge Base

#Content to add first

• Shipping and delivery policies.
• Return, refund, exchange, and warranty policies.
• Payment, gift card, and membership rules.
• Usage, installation, care, and sizing guidance.
• Common pre-sales questions and after-sales procedures.
• Special rules for promotions, pre-orders, or custom products.
• Store-specific instructions that the service team expects HeiChat to follow.

#Available knowledge types

• URL for public help-center or policy pages.
• PDF for manuals, policy documents, and guides.
• TEXT FILE for plain-text reference material.
• PLAIN TEXT for short FAQs, instructions, and policies.
• Q&A for recurring questions that need a controlled answer.

#How to write effective Q&A entries

For one answer, add multiple question variants that express the same customer intent. For a delivery-time answer, this could include:

• “How long does shipping take?”
• “When will my order arrive?”
• “What is the delivery time?”

This helps HeiChat identify the same intent when customers phrase it differently. Keep the answer direct and state important conditions. If the best outcome is to recommend a particular product, use the Product answer type instead of placing an unstructured product name in a text response.

#Update and maintenance rules

Knowledge Base content is a static snapshot from the time it is added. Changes to a source page or document are not automatically synchronized. When content changes:

1. Go to Setup > Knowledge > Knowledge Base.
2. Locate the existing entry.
3. Edit, delete, or re-upload the content as appropriate.
4. Test realistic customer questions again.

When Shopify shipping, refund, privacy, or terms policies change, use Renew Policy to synchronize them. Add an expiration date to temporary promotions, campaign rules, and other short-lived content so HeiChat does not continue using outdated information.

#Why this matters

Knowledge Base is HeiChat's store-specific service memory for store rules, policies, and service procedures beyond the product catalog. It is separate from Learned Products: product information such as titles, descriptions, variants, prices, and inventory is read and synchronized through Learned Products, while shipping, returns, refunds, usage guidance, and other business knowledge are maintained in Knowledge Base. Keep each entry focused on one topic instead of combining unrelated policies in one long document.

---

#6. Enable Gap Detection: find unanswered questions automatically

#Path

Shopify Admin > Apps > HeiChat > Setup > Advanced Settings > Gap Detection

#Recommended Detection Rules

Configure HeiChat to record a knowledge gap when it:

• uses uncertain language such as “I'm not sure” or “I don't have that information”;
• gives a generic answer that does not provide an actionable next step;
• repeatedly directs a customer to support for a question the AI should answer directly;
• asks the customer to rephrase the question multiple times;
• fails to cover a store-specific procedure or policy.

Also define cases that should not be recorded as gaps, such as:

• order status and tracking questions that require real-time account data;
• appropriate refusals of out-of-scope or inappropriate requests;
• refund exceptions or privacy matters that require human review.

#Why this matters

Gap Detection gives HeiChat a form of post-conversation self-review. It does not change the reply shown to the customer; it analyzes whether the conversation exposed missing knowledge. This lets merchants identify recurring weak answers before they become a larger source of frustration or support workload.

Detection Rules are only for classifying gaps. Put tone, recommendation strategy, and reply instructions in AI Response instead.

---

#7. Resolve Knowledge Gaps: turn customer questions into new knowledge

#Path

Shopify Admin > Apps > HeiChat > Setup > Knowledge > Knowledge Gaps

#Recommended process

1. Sort by Hits and start with the questions affecting the most customers.
2. Open a question and read the customer question, AI Response, and AI Insight.
3. Select View Conversation to understand the context before deciding what to change.
4. Decide whether the cause is missing knowledge, outdated knowledge, incomplete product data, or a case that should be escalated.
5. Select Add Knowledge and create a focused Q&A, PLAIN TEXT, URL, PDF, or other appropriate entry.
6. Test the original question and several alternative phrasings in the storefront widget.
7. Select Mark as Resolved after the answer has improved.
8. Use Ignore only when the gap is a false positive or the current AI behavior is already acceptable.

#Why this matters

Knowledge Gaps create a continuous improvement loop:

Review gap → Add Knowledge → Test response → Mark as Resolved

Gaps with high Hits usually represent common customer intents. Addressing frequent questions about shipping, returns, sizing, compatibility, and payment normally improves more conversations than starting with rare edge cases.

Before adding an entry, check whether related information already exists in Knowledge Base. If an answer is outdated or unclear, update the existing entry instead of creating competing content that could lead to inconsistent responses.

---

#8. Use Correction regularly: turn conversation fixes into knowledge

#Path

Shopify Admin > Apps > HeiChat > Conversations > All Conversations

#How to use it

As part of routine operations, review real conversations and AI replies regularly, especially for high-volume topics such as products, shipping, returns, refunds, promotions, and after-sales support. If a reply is inaccurate, incomplete, or inconsistent with your store policy, select Correct this message (shown as Correct in some interfaces) on the relevant AI message.

In the Correction dialog, you can:

• rewrite the customer question so it represents the customer's actual intent more clearly;
• provide a standard answer using Text;
• associate a specific product using Product when that product should be recommended for similar questions.

#Important limitation

Correction is not a mandatory rule and does not permanently override every future reply. It is essentially a shortcut for adding a human correction from a real conversation as reference knowledge for the AI. When similar questions appear, HeiChat may use the corrected answer, but the final response is still generated using the current conversation, existing Knowledge Base content, and other available context.

Use Correction for quick fixes while reviewing conversations. For important, long-term policies or procedures, also organize the corrected information into a formal, clear, and maintainable entry under Setup > Knowledge > Knowledge Base. After correcting a message, test several similar phrasings to decide whether the Knowledge Base or AI Response instructions need further improvement.

#Why this matters

Correction connects “spotting a bad reply” with “adding missing knowledge” without requiring the operator to leave the conversation and recreate the context manually. It lowers maintenance effort and helps HeiChat cover common customer intents more quickly.

---

#9. Configure human escalation: let the AI know when to hand off

#Paths

• Shopify Admin > Apps > HeiChat > Setup > Widget Settings > Live Support
• If your team uses automated ticketing, review Setup > Advanced Settings > Smart Tickets as well.

#What to configure

• Live Support button text.
• Support working hours.
• Holiday schedules.
• Support email and ticket notifications.
• The categories of questions that require human takeover.

#Issues that should usually reach a person

• Requests for a refund exception or policy waiver.
• Damaged, missing, or complaint-related orders.
• Account privacy, payment disputes, or other sensitive information.
• Unconfirmed stock, delivery timing, or order status.
• High-value customers or complex custom requests.

#Why this matters

Strong AI service is not only about answering questions; it is also about knowing when not to guess. A clear human escalation path protects customer trust, avoids unsupported promises, and gives customers a clear next step when AI alone is not appropriate.

---

#10. Test the complete customer journey

After configuration, test HeiChat from a customer's perspective instead of checking only the settings in the admin. Use questions such as:

• “What does your store sell?”
• “Who is this product suitable for?”
• “How long does delivery to my country take?”
• “Can I return this item, and under what conditions?”
• “Do you have a product for this need?”
• “Where is my order?”
• “I need help with a damaged order.”
• The same question in different languages or with different wording.

Check whether the answer uses the correct product and policy information, matches the intended brand voice, makes uncertainty clear instead of guessing, recommends products appropriately, and offers Live Support or another human channel when necessary. Retest questions that previously appeared in Knowledge Gaps.

When an answer is inaccurate, change one factor at a time: check Learned Products, then Knowledge Base, then AI Response and AI Model. This makes the root cause easier to identify.

---

#11. Ongoing maintenance routine

#Weekly

• Review Knowledge Gaps, starting with high-Hits questions.
• Browse a sample of real AI conversations and use Correct to provide a correction when needed.
• Verify that recently added knowledge covers the ways customers actually phrase their questions.
• Look for duplicate or conflicting entries.

#After every policy change

• Update the policy in Shopify.
• Select Renew Policy in Knowledge Base.
• Test related shipping, return, and refund questions.

#After product launches or significant catalog changes

• Confirm that Learned Products has synchronized.
• Test product attributes and recommendations.
• For large-batch product updates, manually select Refresh on the Learned Products page to make sure the changes take effect promptly.

#After a campaign ends

• Review campaign and promotion entries.
• Edit or remove them as needed, or allow entries with an expiration date to expire.
• Prevent HeiChat from continuing to offer outdated promotions.

#Monthly

• Review AI Model performance and token usage.
• Revisit AI Response Custom Instructions.
• Remove outdated Knowledge Base content.
• Review Knowledge Gaps, Support Tickets, Conversations, and Leads to identify the next improvement priorities.

---

#Quick checklist

• Bot name, Store Name, and Store Introduction in Basic Settings are accurate.
• Customer Service Supervisor's Email is monitored.
• Product Learning is complete and includes important products.
• AI Response settings match the brand's voice, length, language, and service boundaries.
• AI Model has been tested with real customer scenarios and balances quality with cost.
• Shipping, returns, refunds, warranty, and other key policies are in Knowledge Base.
• Renew Policy is used after policy changes.
• Temporary information has an expiration date.
• Gap Detection is enabled with useful inclusion and exclusion rules.
• Frequent Knowledge Gaps have been converted into focused knowledge entries.
• Real AI conversations are reviewed regularly, and Correct is used when appropriate.
• Important long-term rules from Correction have been organized into formal Knowledge Base entries.
• Live Support or Smart Tickets provides a clear human escalation path.
• The complete customer journey has been tested from the storefront.