Skip to main content

Common Business Scenarios

This page is not here to teach business users command syntax. It shows how a real business request is usually broken down by an Agent behind the scenes.


Case 1: Sales Follow-up List

A business user might say

Help me find customers created in the last 30 days who still have not placed their first order, group them by industry, and give sales a priority follow-up list. Read-only.

What the Agent will probably do

  • confirm the current business system and customer object
  • locate the customer-related dataset
  • filter customers created in the last 30 days without a first order
  • group the result by industry
  • produce a list the sales team can use directly

What usually gets delivered back

  • a follow-up customer list
  • the potential lead volume by industry
  • a clearer follow-up priority

Case 2: Order Fulfillment Backlog

A business user might say

Help me find paid but unshipped orders from the last 7 days, group them by warehouse, and show me where the backlog mainly sits. Read-only.

What the Agent will probably do

  • pull the exception order details
  • summarize them by warehouse
  • if that is still not enough, run a predefined SQL for the fulfillment chain

What usually gets delivered back

  • an exception-order list
  • a warehouse backlog ranking
  • a practical starting point for operations and warehouse teams

Case 3: Customer Service Response Prep

A business user might say

Use the customer’s mobile number to pull recent orders, tickets, and follow-up results together, then turn them into a response that customer service can use directly. Read-only.

What the Agent will probably do

  • locate the customer by mobile number
  • call a unified snapshot BFF or another existing lookup flow
  • pull the recent tickets, orders, and follow-up data
  • turn the result into a response script or next-step suggestion

What usually gets delivered back

  • a one-page summary of what recently happened with the customer
  • the context most relevant to the current issue
  • a more actionable customer service reply

Case 4: Requirement Data Correction

A business user might say

Help me change requirement 241 to high priority and assign it to Haoying. Preview first, then wait for my confirmation before executing.

What the Agent will probably do

  • read the current record first
  • generate a write preview
  • wait for confirmation
  • then run the real update

What usually gets delivered back

  • a preview of the intended write
  • a controlled and reviewable modification
  • the final confirmation after the update

Case 5: Automated Daily Operations Report

A business user might say

Every morning, generate yesterday’s new-customer count, completed-order count, and refund-order count, then send the result to the operations daily report.

What the Agent will probably do

  • run a predefined reporting SQL or aggregation flow
  • execute it on a schedule
  • output JSON, a table, or a Lark message

What usually gets delivered back

  • a stable daily report result
  • more consistent reporting definitions
  • something that can be forwarded directly to the business team

Which Capability to Use for Which Problem

Business problemWhat the Agent will often choose
lists or detailed recordsdata queries first
summaries or metricsaggregation or SQL
a packaged business resultBFF
notifications, messages, tasksLark CLI, DingTalk CLI, and other ecosystem tools
backfill, correction, deletionpreview first, then write

The two things that matter most from a business point of view

  1. say clearly what business result you want
  2. say clearly whether the request is allowed to write

Once those two points are clear, it becomes much easier for the Agent to choose the right skill, the right CLI, and the right execution order.