Paid Users Exceed 20%: Why E-commerce Merchants Are Flocking to Qwen Office

E-commerce operations have long been dominated by three things: switching back and forth between multiple platforms for data, manually monitoring prices and inventory, and struggling to produce massive amounts of product materials before peak seasons. These tasks require little technical expertise yet consume enormous manpower. A set of data disclosed by Qwen Office CEO Chen Yusen at a product launch event revealed the speed at which AI Agents are being adopted in this field—just over two months after launch, e-commerce has become one of the most deeply engaged industries, with e-commerce merchants accounting for over 20% of paid users.

Multiple merchant cases interviewed by Leiphone.com show that the value of AI office tools in e-commerce scenarios has extended beyond generating a few images or writing a few paragraphs of copy to cross-platform data aggregation, automated inspections, task scheduling, and even business decision support.

Three Grueling Tasks That AI Is Now Taking Over

Multi-Platform Operations: From "Data Archaeology" to Automated Dashboards

Kan Ouli, CEO of Babit Coffee, runs a small team but operates on over a dozen e-commerce platforms simultaneously. In the past, generating a business report required logging into each platform's backend, exporting sales and traffic data, then cleaning, aggregating, and analyzing it. By the time the report was done, the data was often already outdated.

After adopting Qwen Office, the time to generate cross-platform business dashboards was compressed to 30 minutes. For events like "Double Eleven" that require rapid decision-making, improved data timeliness directly affects advertising and inventory decisions.

Price Monitoring: From Manual Checks to Scheduled Daily Reports

Xiaoyu, an operations staff member at Bear Electric, spends her days repeatedly checking product prices on Tmall and JD.com. Each check takes 15 to 30 minutes, and she has to calculate coupons and platform subsidies into actual out-of-pocket prices—missing short-term price drops or product delistings if she's not careful.

Qwen Office's approach is to automatically collect out-of-pocket prices from both platforms at 8:30 AM and 6:00 PM daily, generating a price report sent to email. The report includes not only price changes for the company's own products and competitors, but also inventory status, 7-day trends, and price differentials. Anomalies can be spotted faster, and staff only need to handle the parts requiring judgment.

Material Production: From Four Months to Fifteen Days

Sandy runs two Tmall lingerie flagship stores and has 15 years of e-commerce experience. She did the math: during the autumn/winter new season, the two stores need nearly 10,000 seeding images. Under the traditional model, a designer producing 100 images a day is already at capacity—10,000 assets would require about 100 working days, or four to five months end to end.

"Being slow means missing the entire peak season."

Sandy's solution wasn't to treat Qwen Office as a simple image generation tool, but as a task scheduling layer—using connectors to call third-party image generation software, with Qwen Office organizing tasks, passing instructions, and driving batch execution. The original four-month material preparation cycle was compressed to about 15 days. In one actual generation run, 376 images were usable and 72 were discarded, achieving an 84% usability rate.

The Role of Connectors and Task Scheduling Layers in Agent Architecture

A Connector is a standardized interface for establishing communication between an Agent and external tools or platforms—essentially a "translation layer" that allows the Agent to call third-party software functions without manual user operation. The task scheduling layer is responsible for breaking down complex goals into executable subtasks and organizing their execution order based on dependencies. In Sandy's case, Qwen Office doesn't directly generate images but calls professional image generation software through connectors while also serving as the task orchestrator—deciding when to call, what instructions to pass, and how to filter results. This architecture enables the Agent to integrate the capabilities of multiple specialized tools to complete complex workflows that no single tool could handle alone, and explains why the value of AI office tools is shifting from "replacing a specific step" to "orchestrating the entire process."

When AI Gets More Capable, What Do Merchants Do

After AI takes over repetitive labor, merchants' time allocation has shifted noticeably. Liu Xianglong now spends only 2 to 3 hours a day on operations work, using the saved energy to determine business direction, manage AI, and oversee products and supply chains. He also converted his personal experience into prompts and built a three-tier dashboard system covering platform, store, and product levels.

Kan Ouli uses AI for market change analysis to help determine the direction of new coffee product development. "No matter how deep one person's experience is, it's only experience from their own circle. In the past, developing new products was about what I wanted to sell; now I use AI to first see where the market is heading."

Sandy's summary of AI's role: "AI is the hands, I am the brain." AI can draw over a thousand variations a day, but what counts as a good image, which scenarios to scale up, and which to stop—these require human definition. Her team places high-performing assets into a premium asset pool, writes effective judgment criteria into a rules library, and then combines them with advertising performance to decide whether to double down or adjust, forming a cycle of data analysis, performance validation, and experience accumulation.

Two Conditions for Agent Adoption in E-commerce

The Agent wave's rapid entry into e-commerce is directly related to industry characteristics. E-commerce is close to transactions, so AI's effects can be directly reflected in sales, conversion rates, and costs—giving merchants clear motivation to try. At the same time, the e-commerce industry's years of digital accumulation—management platforms, SaaS software, data analysis tools—provides the accessible data foundation for Agents.

But shortcomings exist as well. Tools across platforms and stages are fragmented, and the intelligence level of vertical platforms and SaaS is limited, causing merchants to hesitate when choosing AI tools.

Qwen Office E-commerce Edition's approach is to build an independent third-party e-commerce Agent, open to all e-commerce platforms and merchants, with merchant tools and industry software from various platforms able to connect. Chen Yusen particularly emphasized the importance of Context: what truly understands a merchant's business is the data hidden in daily operational behaviors—platform monitoring data, strategies in business documents, internal communication records, and instructions from interactions with the Agent.

The path being explored by the industry is: first establish a governable enterprise context layer, then organize digital employees to collaborate on top of it. Qwen Office E-commerce Edition is also rolling out proactive operational capabilities, letting the Agent proactively discover anomalies in inventory, advertising, conversion, after-sales, and other areas. For small and medium merchants lacking programming capabilities and with limited staff, this lowers the barrier to using AI.

What Context Means in the Agent Sense

In the AI Agent field, Context doesn't broadly refer to conversation history, but rather the complete set of information an Agent can access when executing tasks, including system prompts, tool return values, business data, historical interaction records, and more. The quality and richness of context directly determines whether an Agent can make judgments that align with business reality. In e-commerce scenarios, platform operational data, strategies in business documents, and internal communication records constitute the context specific to that merchant. This information is often scattered across different systems and needs to be governed—cleaned, structured, and access-controlled—before it can be effectively used by the Agent. The industry's proposed "governable enterprise context" is essentially building a trustworthy data foundation for the Agent, so its decisions are grounded rather than generated out of thin air.

The Efficiency Gap May Be Reopened

E-commerce competition has largely been a competition of execution capability: whoever can launch new products faster, adjust prices more frequently, produce more materials, and monitor more channels is more likely to capture opportunities.

As AI begins taking over part of the execution work, the efficiency gap between merchants may be reopened. Beyond execution, the weighting of industry understanding, market judgment, and control over products and supply chains will rise.

Kan Ouli positions Babit as "a specialty coffee e-commerce company transforming itself with AI." Before entering e-commerce, he was a musician who had never written code. Now, AI office tools that are ready to use and operable by ordinary employees allow him to free up more energy to think about business direction and strategic decisions.

"The times are shaking so much—how do you avoid being thrown to the ground." In his view, with AI's technological foundation, merchants need to re-examine their ways of working: which tasks actually don't need to be done by "humans" anymore.

Background

The Difference Between AI Agents and Traditional Automation Tools

Traditional automation tools (like RPA) rely on preset rules and can only execute fixed steps, easily failing when encountering anomalies or unstructured input. AI Agents, by contrast, possess a degree of reasoning and planning capability, able to autonomously break down tasks, select tools, and process intermediate results based on goals. For example, in price monitoring scenarios, RPA can scrape page prices on a schedule but cannot understand the out-of-pocket price after coupon stacking, nor judge whether a price differential anomaly warrants an alert. Agents can understand semantics by calling large models and process numerical values by calling calculation tools, completing more complex judgment chains. This is why the article emphasizes that AI's value has extended from "generating a few images" to "business decision support"—the core difference being that Agents can handle unstructured information and make adaptive adjustments in dynamic environments.

One-line Summary

Two months after Qwen Office E-commerce Edition launched, AI Agents are taking over repetitive e-commerce work, and the efficiency gap may be reopening.

Paragraph Summary

Through merchant cases including Babit Coffee, Bear Electric, and Sandy, the article demonstrates AI Agent results in e-commerce: cross-platform data aggregation went from manual "archaeology" to 30-minute automated dashboards, price monitoring from 15-30 minute manual checks to twice-daily scheduled reports, and material production from four months compressed to 15 days with an 84% usability rate. With AI handling execution, merchants redirect time toward business judgment, AI management, and supply chain oversight. The article notes that Agent adoption in e-commerce depends on two conditions: directly quantifiable results that drive trials, and the industry's digital accumulation providing a data foundation. It also emphasizes the importance of Context governance and predicts that once execution is no longer a barrier, the weighting of industry understanding and market judgment will rise.