Vendor Case: How a Factory Successor Tested Feishu and AI on the Shop Floor

On Feishu, management can begin by migrating their operations onto the platform—one table, one group, or one real-world problem at a time—shaping it precisely to fit their company’s unique structure.

Xiao Pan aims to change more than just the tools used in a factory. He wants data scattered across paper forms, chat groups, and personal phones to return to the enterprise; he wants the expertise held within employees’ minds to remain within the organization; and he hopes tasks that once relied on managerial coordination will instead be driven by facts and standards.

In July, a young man wearing a white safety helmet took the stage at the Feishu AI Mastery Conference in Beijing.

At the event, Xiao Pan said the factory had piloted AI-assisted analysis of camera images to flag helmet use on a production line. This is an interviewee account. Any use of worker images requires a lawful basis, clear notice, data minimisation, access controls, retention limits and a defined human response process.

The young man on stage is Xiao Pan, who describes himself as part of the post-00s generation of factory successors. His family’s business—Wuhu Hoton Automotive Industry Co., Ltd. (hereinafter "Hoton Automotive")—by his account has operated for years and primarily makes automotive chassis components such as control arms and bushings.

Inside the factory, people call him "Little Boss," but Xiao Pan sees his role more as "the factory’s only FDE"—someone who must understand the business while personally implementing technology on-site.

His AI transformation didn’t start with a grand strategy, but from a challenging homecoming and succession process.

First, Consolidate Operational Data

In January 2025, Xiao Pan returned to his family’s factory. Before arriving, he expected the transition wouldn’t be too difficult. But upon stepping inside, his feeling could be summed up in seven words: "One look, and the sky had fallen."

Part of the company’s information flow still relied on paper documents. Excel automation depended on legacy VBA macros, and daily communication happened through WeChat groups—numerous groups where files expired after seven days, and work messages were tangled with personal chats. The company’s use of AI was also in its early stages, mostly limited to casual conversations and information queries.

Yet as competition in the auto parts industry intensified, businesses needed finer cost control and operational management. Any equipment downtime, material delay, or miscommunication could directly impact delivery schedules and performance. The limitations of the old model began to surface: without data accumulation, not only was there no foundation for AI, but problems were hard to trace accurately—everything had to be manually filled out.

Under this pressure, after joining the factory, Xiao Pan gradually shifted from sales to operations management, taking on more cross-departmental coordination. Different roles often interpreted the same matter differently, forcing him to constantly align field information, departmental goals, and management expectations.

A boss might resolve one dispute, but cannot stay forever in every group, every spreadsheet, and every workshop. Xiao Pan gradually realized that what truly needed changing wasn't just the form of communication—but what the company relied on to communicate: memory, managerial decree, or data and standards speaking for themselves.

In September 2025, Xiao Pan said he began introducing Feishu to staff with collaboration needs. His internal pitch was simple: “Let’s consolidate our organisational data.”

“They understood that data has value,” said Xiao Pan. “So I used that as the entry point—data should be consolidated and stored in one place.”

Rolling out Feishu was not easy. New software meant new habits, so Xiao Pan taught as he went. Because “consolidating organisational data” sounded abstract, he started with visible, concrete problems.

So, Xiao Pan decided to start with a single piece of paper.

Inside Hoton Automotive’s factory

Starting with a Single Sheet of Paper

Sales orders, purchase receipts, daily production reports, warehouse ledgers, equipment inspections—within a short time, Xiao Pan and an IT colleague migrated many of the factory’s paper forms into Feishu’s multi-dimensional tables.

Equipment inspection was a small yet highly representative example.

Previously, each machine had a paper checklist beside it. Workers would record checks manually, often logging everything into Excel only during lunch or after work. "If all records are on paper, they’re practically worthless. I can’t verify whether inspections actually happened, nor assess their quality," Xiao Pan said.

According to the case, workers could scan QR codes, take photos and submit reports on-site, while management gained one view of the workstation, machine and reported issue. The configured workflow could route exceptions to team leaders and then the equipment department; timing, accuracy and access controls should be tested.

When the three tables—field inspections, equipment repair work orders, and spare parts inventory—were linked, inspections gained real value. Previously isolated information began forming a complete picture: which machines repeatedly failed, which repairs took too long, and which non-standard spare parts were unexpectedly out of stock. Whether management can act before breakdowns, not after, is still being judged against repair records.

Daily production reports followed the same path. In the past, workshop leaders submitted Excel sheets to planners for compilation. Xiao Pan said multiple workshops later entered data in Feishu’s multi-dimensional tables and his configured workflow could generate management reports. Buyers should verify the actual coverage, update frequency and review process in a pilot.

"In Xiao Pan’s view, multi-dimensional tables function as the factory’s database—the foundation his agents are built on."

For this factory, only when business data returns to the organization can AI understand what’s happening—and potentially participate in operations, rather than remain detached from them.

Forging equipment ledger in the multi-dimensional table (partial)

When AI Entered the Work Group

After bringing data back via multi-dimensional tables, Xiao Pan began embedding intelligent agents into daily collaboration.

Initially, management still saw AI as "just a chatbot you can ask questions." Xiao Pan needed a scenario where they could directly perceive AI’s practical impact.

He created a group message dashboard.

He named the AI assistant "Xiao He," derived from the company name "Hoton," and added it to key Feishu groups. As configured by Xiao Pan, it captures events in the chat, flags deadlines, responsible persons, and pending tasks, and summarizes these at intervals into multi-dimensional tables—behaviour a buyer should confirm in their own pilot groups.

In the past, when a leader assigned a task in a group chat, if the recipient failed to record it promptly, the message could easily get buried. Days later, follow-ups revealed nothing but confusion. "In my experience, the group message dashboard addresses this—listing who told whom what, who’s responsible, and what’s next—all publicly visible, as we’ve configured it."

Xiao He’s automated group message tracking dashboard

This was the first time management directly experienced AI’s value. It didn’t replace anyone’s job or invent a new process—it simply preserved information that would otherwise be lost, ensuring accountability didn’t sink with chat history.

If the message dashboard won management over to AI, then the "Quality Expert" agent marked AI’s first intervention in interdepartmental collaboration challenges.

Hoton Automotive was upgrading its quality management system. Its IATF 16949 documentation needed reorganization, but progress stalled—the quality department knew the standards, while operational departments understood actual workflows. They disagreed on who should draft the documents, causing deadlock.

At this impasse, Xiao Pan decided to become the "one who understands the standards." During his vacation, he self-studied the 16949 standard, consolidating notes, course materials, and system documents into Feishu’s knowledge base, building the "Quality Expert" agent. He set strict rules: "Any output must cite supporting clauses—exactly which standard, which clause, which sentence—and show the original text."

His analogy: "The quality system is like internal corporate law. When people don’t fully understand the law, you bring in a lawyer."

Once, an operational department submitted a draft document for review in a group chat. Instead of commenting directly, Xiao Pan "@'d Quality Expert" to have it evaluate the submission against the standard.

In Xiao Pan’s account, the agent listed possible improvements, cited clauses and drafted revision suggestions, giving the team a shared checklist. IATF 16949 and other quality-system requirements still need qualified staff to check the current official standard and site evidence; model output is not a compliance conclusion.

"Once the agent laid out the criteria clause by clause, the conversation shifted from role-based stances back to the standards themselves—making acceptance much easier." But Xiao Pan didn’t want an "arbiter" in every step. "I hope AI directly optimizes your output." Over time, expert staff’s corrections and supplements to the agent’s responses gradually became reusable organizational knowledge.

"What the Boss Delivers Must Be Ready-to-Use"

However, showing management that AI is useful only answers "Is it worth using?" To make it part of daily work, the question becomes "Will employees know how and want to use it?" Xiao Pan didn’t ask staff to learn multi-dimensional tables, write prompts, or build agents. Instead, he and the IT colleague prepared the tools first, then handed them ready-made to users.

"What the boss gives must be ready-to-use—activatable with a single sentence," said Xiao Pan. "If you ask them to learn it, one, they’re already busy; two, for an organization our size, the friction cost is too high. Our approach is to build it right and deliver it to you."

Thus, Xiao Pan designed lightweight systems in Feishu multi-dimensional tables that mirrored existing workflows, preserving familiar ways of working. Where they once entered numbers in Excel, they now do so in mobile multi-dimensional tables. Employees don’t need to relearn entire processes—only where to input data, what results to check, and what actions to monitor. "Our company looked a certain way before—on multi-dimensional tables, it looks basically the same."

In Xiao Pan’s case, the team started with one table, one group and one operational problem instead of replacing every process at once. This does not establish product fit for every factory; buyers should compare existing systems, integration effort, permissions, export and exit arrangements.

Only when data stays within the organization, group chats generate traceable tasks, and knowledge bases absorb expert experience, can AI evolve from a personal tool into an integral part of the organization.

When asked why he’s willing to invest so much time in these difficult changes, Xiao Pan says it ultimately comes down to a sense of responsibility.

This sense of responsibility manifests in concrete actions: from writing Excel macros, to pushing Feishu adoption, to building tables and AI applications with an IT colleague, and teaching employees step by step.

That’s why, on stage, he didn’t start with lofty AI concepts, but with a safety helmet. Because to him, AI transformation must answer one question: Have real problems on the shop floor actually been solved?

For Hoton Automotive, AI transformation isn’t some distant future. It’s one less sheet of paper beside a machine, one task in a group chat that didn’t get buried again, one document smoothly completed across departments, one piece of expertise retained in the company after a specialist answered a question.

What Xiao Pan truly wants to bring onto Feishu has never been just factory data—but the company’s ability to keep moving forward. And he wants this story seen—not just for Hoton Automotive.

"Companies like ours are actually the majority in China," he said. "But in the conversation about AI transformation, voices like ours are rarely heard."