User ManualOPERATIONS FIELD GUIDE
User Manual/Shared setup

Shared setup

Complete Shop Work in Chat

Set up Customer Service, check Shop status, and complete business tasks from Desktop or a connected message channel.

#Tell Chat what you want to accomplish

Chat is a shared starting point for Customer Service, Affiliate, and day-to-day Shop work. You do not need to find every setting first or understand how the system works behind the scenes. Name the Shop, describe the outcome, and state any limits. The Agent will review the current situation before returning a result or recommending the next step.

APP CAPTUREThe English light-theme Chat workspace with session tabs, runtime controls, and composer.

Chat is especially useful when a task requires information from several places, a comparison across Shops, or analysis before a change. For an important update, ask the Agent to show its proposal and wait for approval before saving.

#Chat from more than the Desktop page

The Chat page is the full workspace, but it is not the only way to talk to the Agent. Once connected and authorized, Feishu or Lark, WeChat, Telegram, and other messaging channels can also send messages directly to the Agent and start a new conversation.

At the office, use Desktop Chat to review the full context, attach files, and handle longer tasks. When you are away, use your phone to ask about Shop status, summarize exceptions, or start a new check. For example:

Check our three Shops for unusual orders today and list anything that needs my decision.

Summarize today's Customer Service escalations and pending Affiliate tasks. Do not change any status.

Review Brightloom US Customer Service performance for the last 24 hours and tell me the three issues that need the most attention.

The Agent can work with Shops and accounts that have already been connected. For buyer replies, setting changes, or other important actions, include “ask me before proceeding.” Each entry point starts and retains its own conversation. If earlier analysis happened elsewhere, include the necessary Shop, date range, and task context in the new message.

Go to Message channels to connect and test Feishu or Lark, WeChat, or Telegram. On first use, you may need to send an initial message and complete pairing or allowlist setup. To remain available, the Desktop responsible for those channels must stay online with a working model connection.

#Set up a Customer Service prompt in Chat

For example, send:

Set up the AI Customer Service prompt for Brightloom US. Review its products, the last 30 days of Customer Service conversations, and the existing settings first. Show me your recommendation for approval before saving anything.

The Agent will:

  1. confirm which Shop you mean;
  2. review the current Customer Service settings, key products, and recent conversations;
  3. identify recurring shipping, product, and after-sales issues;
  4. organize what AI may handle, what needs a human decision, and what it must not promise;
  5. show you the proposed changes and wait for approval before saving;
  6. check the settings again after saving and report the result.

If one Shop already has a mature setup, you can also say:

Use the Brightloom US Customer Service rules as a starting point for Brightloom UK. Keep the general after-sales principles, but adapt the prompt to the UK catalog, shipping, and policies. Do not save it yet.

When a task requires local-device assignment, clearing an existing setting, or reviewing the full form, the Agent will direct you to Shop Management or the Customer Service configuration page.

#What else you can do in Chat

Chat is also useful for:

  • Customer Service analysis: summarize recent conversations, identify frequent issues, and find gaps in the current instructions;
  • Business analysis: compare performance by Shop, market, and date range, then highlight the largest changes;
  • Daily investigation: look into unusual orders, fulfillment, products, inventory, or warehouse issues;
  • Affiliate follow-up: review Creators, Samples, and Collaborations, then summarize what needs attention;
  • Cross-Shop checks: find incomplete setup, unusual performance, or Shops that require human attention;
  • Proposal-first changes: ask the Agent to prepare a recommendation and execute it only after approval.

Whether a task can be completed depends on the connected Shops, accounts, and services. If something is not ready, the Agent will explain what is missing and where to complete the setup.

#Describe the task clearly

A useful request states the goal, the Shop or date range, what must not happen, and the result you want.

Analyze without changing anything:

Review the last seven days of Customer Service conversations for Brightloom US. Group them into shipping, product, and refund issues. Do not change settings or reply to buyers. Tell me which policies are missing from the current prompt.

Require approval before a change:

Draft a revised Customer Service prompt from that analysis. Show me the main differences and do not save yet. After I approve, update the Shop and check the saved result.

Compare business performance:

Compare SPS and Customer Service escalation rate for our three US Shops over the last 14 days. Return a table with the largest changes, likely causes, and the next checks to run.

If the scope is unclear, the Agent will first confirm the Shop, time range, or permitted action. For buyer replies, setting changes, or other important actions, include “ask me before proceeding.”

#When to use a dedicated page

Chat works best for completing a connected series of steps around one goal. Dedicated pages are better for scanning large amounts of information and working through a fixed queue.

Open the corresponding page when you need to:

  • review many Customer Service conversations, Affiliate tasks, tables, or charts;
  • make human-review decisions one item at a time;
  • bind a local device, connect an account, or maintain form-only settings;
  • compare a large amount of visual information during routine operations.

A common workflow is to analyze the issue and prepare a plan in Chat, then use the business page to inspect the queue, status, and final result.

#Manage sessions and context

Each session tab keeps one workstream separate. Keep a multi-step task in the same session so the Agent retains the goal, analysis, and approval history. Start a new session when switching to another Shop or an unrelated task so Shops and date ranges do not become mixed.

Before sending a task, confirm that the current model and connection state are available. You can also attach files or images for the Agent to read reports, screenshots, or other source material.

#Turn a proven workflow into a schedule

Develop and refine the task in Chat until its input and output are reliable. When the same work should run daily, weekly, or at a fixed time, turn it into a Scheduled Task.

TK COPILOT · OPERATIONS FIELD GUIDE© 2026 RIVON LLC