Kidswant has developed more than 300 business AI agents over the past year, covering over 400 scenarios, with nearly RMB 3 billion in sales coming through AI-led marketing workflows. These figures are worth noting: the competitive center of enterprise AI is shifting from “connecting to models” to “embedding AI in business operations.”

Kidswant is a nationwide maternal and child care retailer with stores across nearly 300 cities, serving close to 90 million users. Its agent platform is built on Alibaba Cloud’s DataWorks Data Agent and covers three areas: store managers, employees and precision marketing. A Data Agent can be understood as an intelligent assistant that helps business teams access enterprise data, generate analyses and execute tasks.

What this is

Kidswant previously faced three major challenges: wide variation in store managers’ capabilities, difficulty ensuring consistent service from roughly 10,000 frontline consultants, and inconsistent definitions across multiple data systems accumulated over more than a decade. Its AI store manager assistant combines inventory, orders, membership and trend data to recommend operational priorities for different stores; the employee assistant identifies membership opportunities and produces communication references; and precision marketing connects audience identification, outreach, content generation and transaction conversion.

These applications are supported by a unified data and agent foundation. DataWorks Data Agent provides data integration, governance, analytics, security auditing and skill development, while the Context Graph semantic knowledge graph standardizes metric definitions and data relationships. The addition of MCP (Model Context Protocol, a standard protocol that lets AI connect to external tools and data sources) makes it easier for enterprises to integrate existing systems with agents.

Industry view

This is more of an enterprise case moving from pilot projects into everyday operations than a conventional AI product launch. Kidswant reports that stockout rates for best-selling products fell 1.3%, shipping fulfillment rose to 99%, and revenue per customer increased by nearly 6.6%. These results suggest that data standardization and clarity of business rules are often closer to real-world performance than model parameter size.

The risks are equally clear. Most of these metrics were disclosed by the company itself, without control groups, consistent cost definitions or long observation periods, making it difficult to conclude that the agents caused every improvement. Agents that generate incorrect business advice, rely on outdated labels or have improperly configured permissions can also scale errors. Moreover, RMB 3 billion in “AI-led sales” does not mean RMB 3 billion in net profit; human review, system costs and actual conversion rates still need to be assessed separately. What the industry must validate is not a one-off demonstration, but whether agents can consistently and compliantly create incremental value after months of operation.

Impact on regular people

For enterprise IT: Enterprise AI procurement will shift from chat assistants toward data connectivity, permission governance and workflow orchestration. The CTO’s central task will be to turn fragmented business experience into executable rules and skills.

For individual careers: Store managers, sales staff and customer service teams will first receive automatically generated customer lists, operational recommendations and draft scripts. Repetitive judgment work will decline, but final communication and critical decisions will still require human experience and clear accountability.

For the consumer market: Recommendations in maternal and child retail will become more granular and personalized, potentially relying more heavily on user profiles. Consumers will receive services that better match their needs, while paying attention to the transparency of personalized recommendations and the boundaries of data use.