From collaboration tools to digital employees, LINLEE explores how AI can integrate into operations, transform roles, and reshape organizations through Feishu.
From collaboration tools to digital employees, LINLEE leverages Feishu to examine how AI permeates business, reshapes positions, and redefines organizational structures.
In recent years, China’s new-style tea market has seen rapid store expansion, price competition, frequent launches and co-branded campaigns.
Founded in 2012 in Zhanjiang, Guangdong, LINLEE chose fragrant lemons as its signature ingredient to popularize the hand-squeezed lemon tea—a drink beloved across Guangdong, Hong Kong, and Macao—into a broader national market. Its iconic fluorescent green color scheme and the yellow duck giveaways included with drinks have created a strong and memorable brand identity. Buyers should verify LINLEE's current store count with the company.
As it evolved from a regional brand into a national chain, new challenges emerged. Beyond ensuring consistently fresh and delicious lemon tea, the company needed to maintain efficient operations across its geographically dispersed, predominantly franchise-based store network.
This is when AI entered LINLEE’s agenda. Initially perceived as a new tool for boosting efficiency, it soon raised deeper questions: How can a company truly embed AI into real business processes and ultimately transform it into organizational capability?
Who Owns the AI Mission?
In December 2024, Chen Bao, then head of LINLEE’s IT department, received an order to implement Feishu.
AI? That would come later.
LINLEE’s motivation for adopting Feishu was straightforward—to solve collaboration challenges. Chen Bao gave an example: a menu spreadsheet required confirmation from several people. After Person A made changes, the versions held by Persons B, C, and D remained outdated, making it unclear whose version was final and often requiring another round of verification.
According to the team, moving information online enabled multi-user collaboration and configured reminders and alerts in Feishu’s multi-dimensional tables. Which tasks can be completed without code should be tested against the current version and tenant configuration.
LINLEE co-founder Chen Xu’an vividly remembers his first experience using AI-powered “auto-fill text” in multi-dimensional tables. He and Chen Bao immediately compiled a long list: contract review, information consolidation, data extraction—could all these tasks be entrusted to AI?
By then, AI had become an unavoidable topic for enterprises. Chen Xu’an had spoken with many business owners and found a shared sentiment: everyone wanted to embrace AI, but no one had seen clear success or knew where to begin. Seeking support from the IT department often led to lengthy and costly system projects; letting employees experiment independently rarely translated individual interest into company-wide capabilities.
The LINLEE team quickly realized that the real challenge wasn’t “whether there were AI tools,” but rather how to transform AI from individual experimentation into a sustained capability serving real business needs.
LINLEE’s lemon tea
Thus, during a management meeting, LINLEE’s CEO proposed an AI-driven transformation, assigning leadership to the General Office Process Optimization Team, responsible for coordinating process management and AI implementation.
The Process Optimization Team could recruit talent across departments, access data, and standardize workflows. When AI touched existing roles and processes, the team had the authority to coordinate adjustments involving business, technology, and HR functions. In December 2025, Chen Bao transitioned to lead the General Office Process Optimization Team.
Like many companies actively embracing AI, LINLEE initially took some missteps. At first, the team aimed to directly quantify the labor-efficiency value brought by AI: estimating potential staffing optimizations from improved efficiency and considering allocating part of the labor savings as team incentives.
But as implementation progressed, Chen Xu’an noticed a shift in focus. The conversation drifted from “Can AI do the job better?” toward how labor savings and incentives would be distributed.
The initiative was quickly halted, leading LINLEE to establish a principle: AI implementation must not start by reverse-engineering labor-efficiency outcomes, but should instead originate from solving real business problems.
First, identify specific tasks and let AI stably take them over, allowing the business to see tangible results. Changes in role division and organizational structure should follow only after business value has been proven—not predetermined.
With direction clarified, new challenges arose: more and more departments wanted to adopt AI, yet the Process Team lacked sufficient manpower. Assigning a dedicated AI specialist to each department was impractical.
To address this, Chen Bao designed a “co-building” AI growth mechanism: if a business department wanted long-term, dedicated AI support, they would allocate one full-time position to the Process Team. The team would then centrally hire and train an “AI Growth Officer” to work closely with that department over the long term.
Initially, the Process Team had only four core staff members. Over time, more business units willingly contributed a position in exchange for a dedicated AI Growth Officer.
This arrangement held both sides accountable. Business departments, having committed a headcount, could no longer treat AI as just a slogan. The Process Team, receiving the headcount, could not deliver merely flashy demos—they had to build and sustain actual AI applications.
Chen Xu’an’s view was clear: if AI remains at the individual usage level, it’s still just a tool. Only when it genuinely takes on part of the work does it begin to embody the meaning of a “digital employee.”
With people driving AI forward and a methodology for integrating AI into the organization becoming clearer, LINLEE now needed to find a sufficiently real-world role to validate whether this approach actually worked. They chose Operations Supervisors.
“Duck Duck” Joins the Team
LINLEE named this digital supervisor “Duck Duck,” inspired by the brand’s famous yellow rubber duck.
Why operations supervisors? Customer service and design tasks might more easily demonstrate AI impact, but supervisors form the largest and most operationally critical team, bridging headquarters, franchisees, and stores. If this role couldn’t be transformed, a digital employee would remain nothing more than a showcase demo.
In Chen Bao’s view, supervisors should act as “business experts” for stores, identifying issues and offering growth recommendations. But in reality, they were often bogged down by repetitive and complex tasks.
In summer 2025, Chen Bao and his team traveled across major regions, shadowing supervisors store by store, doing exactly what supervisors did.
Before field observation, Chen Bao’s impression of supervisors was limited to “long store visits.” Only after experiencing it firsthand did he realize that most of the three-hour visit was spent retrieving data—operational metrics, procurement records, historical inspection results—scattered across different systems, with target achievement rates needing manual calculation. Only after all this was done could they finally discuss operations with franchisees.
Chen Bao wanted AI to take over those repetitive yet indispensable tasks that came before judgment and decision-making. But the first version missed the mark.
In October 2025, the team built a conversational Q&A tool: users typed questions into a chat box, and AI returned data. It resembled an AI customer service bot but failed to integrate into supervisors’ daily workflows. After less than half a month of internal testing, Chen Bao scrapped the first-generation “Duck Duck” and started over.
According to LINLEE, the redesigned second-generation “Duck Duck” entered pilot use in North and South China in February 2026. The team rebuilt it around the supervisor workflow and used Feishu aPaaS to bring operational, procurement, channel-score, task and OA information into one workspace. The current rollout scale and integration status should be confirmed with the company.
Interface of “Duck Duck”
Although data was now centralized, Chen Bao remained unsatisfied. This version of “Duck Duck” still felt like a passive business tool awaiting queries, not a collaborative colleague.
LINLEE says the team then used Doubao Workmate on Feishu to add an agent interface, natural-language interaction and proactive updates within the Feishu workspace. Actual functions, accuracy, permissions and coverage should be tested in the current tenant.
As described by LINLEE, “Duck Duck” is intended to help supervisors check their own and subordinates’ KPIs, generate structured daily reports, answer SOP questions, and track assignments, while for store-facing tasks it is designed to aggregate operational data, identify weak spots, and prepare materials for store visits. Actual performance should be validated in a pilot before commitment.
According to LINLEE, a third-generation “Duck Duck” is being rolled out across its supervisor teams nationwide. Chen Bao’s team has embarked on another round of nationwide store visits, gathering feedback from supervisors and franchisees to determine what tasks “Duck Duck” should take on next.
Digital employees aren’t designed once and done—they are continuously redefined through real-world use.
From One Digital Employee
To an Organizational Transformation
According to the LINLEE team, the clearest early signal from “Duck Duck” was increased store coverage per supervisor, though this remains the company's own account.
Chen Bao pointed to the team's own operational baseline: stores managed per supervisor, tracked month by month from March to July this year, reportedly rose steadily—an internal figure that outside buyers could not independently verify. For any chain weighing a similar tool, the starting question is whether that same coverage metric can be measured reliably today.
This figure cannot be entirely attributed to “Duck Duck,” and LINLEE's experience does not by itself establish that automating repetitive tasks expands individual management capacity—though it is consistent with that hypothesis.
The ripple effects quickly spread beyond the supervisor role. The finance department requested a data-analysis-focused digital employee, design teams sought AI assistance for basic graphic creation, and recruitment began exploring resume screening and interview support.
Frontline teams were no longer just waiting for solutions from headquarters. Some supervisors personally used Doubao Workmate to build custom agents, experimenting with daily performance briefings sent to their assigned stores to meet personalized work needs.
At an event hosted by the Feishu Enterprise AI Architect Training Program, Chen Bao shared that LINLEE has deployed a growing portfolio of AI applications. More significant than the number is the shift in engagement: business units are now raising problems, participating in development, and providing continuous feedback. Conversations have evolved from “What AI tools should we buy?” to “Which responsibilities can be assigned to digital employees?”
According to Feishu, the Enterprise AI Architect Training Program is an AI talent development initiative for senior managers at client companies, intended to support enterprise-wide AI adoption; details and outcomes should be confirmed with the vendor.
“Duck Duck” has also prompted LINLEE to rethink scalability in chain operations. Traditionally, expanding store count meant increasing headcount proportionally.
“Duck Duck” Store Inspection Dashboard
Chen Bao noted that if store expansion continues, relying solely on hiring more people to support management would lead to ever-rising costs. What “Duck Duck” aims to achieve is shifting from “people searching for information” to “AI connecting information flows,” freeing people from constant data gathering and alignment so they can focus on communication, judgment, and decision-making.
This reflects Chen Xu’an’s thinking on current management practices. In the past, to ensure execution quality, organizations invested heavily in process verification. Now, as AI increasingly handles standardized and repetitive tasks, teams can reduce redundant checks and redirect energy toward outcome evaluation, business judgment, and strategic decisions—establishing a clearer human-machine division of labor.
As AI integration deepens, new types of work have emerged within business teams: defining what agents should accomplish, continuously adjusting requirements based on changing business needs, and reviewing final outputs. LINLEE refers to such roles as Agent Owners.
This reflects LINLEE’s understanding of digital employees: it’s not simply adding an AI tool, nor merely replacing a role, but fundamentally rethinking how different tasks within a role should be distributed.
Rather than predefining “how many jobs AI will replace,” LINLEE focuses first on verifying whether AI can reliably handle specific tasks, then determining role evolution based on actual outcomes. While some repetitive duties may decrease, new responsibilities for managing digital employees will emerge.
“Duck Duck” is one LINLEE case and does not show that building an agent completes an AI transformation. LINLEE treats it as a starting point for observing changes in roles, management and organisational boundaries; sustained use and operational data are still needed.
In Chen Bao’s view, AI doesn’t automatically make an organization stronger. Instead, it “amplifies clarity in processes, reliability in data, and accountability in responsibilities,” while exposing previously unarticulated problems. “Duck Duck” is merely the first “employee” to join. The real change lies in LINLEE beginning to understand roles, processes, and organizational capabilities in an entirely new way.