Chapter 1: The Four Movements of Collaborative Work and Real Business Pain Points
Last month, I visited a manufacturing enterprise in the Yangtze River Delta region. Standing in a roaring workshop, I stared at a digital dashboard updating in real time—and suddenly it clicked: the collaborative work we passionately debate in office buildings should have long been interwoven with such authentic, frontline business flows.
Looking back at the evolution of collaborative work, the trajectory is actually quite clear. It began with “communication tools,” designed to ensure information reach; then evolved into “process systems,” digitizing workflows like approvals and calendars; next came “point AI tools” that help draft copy or generate summaries—addressing isolated efficiency bottlenecks. Today, the industry is calling for “organizational intelligence,” where AI integrates deeply with core business operations.
Yet reality often falls short. The most frustrating pain point today is that many AI tools remain stuck at the “point solution” stage. They operate outside genuine business contexts and have never grappled with enterprises’ proprietary data—leaving them little more than soulless chatbox toys.
To understand the value of a collaborative foundation, you cannot judge solely by which tools it offers—you must examine what authentic scenarios it has internalized. Lark delivers foundational capabilities familiar to all users: instant messaging, multi-dimensional spreadsheets, documents, calendar, video conferencing, and approvals. But these are merely the bedrock. The true differentiator lies in how third-party AI assistants or intelligent work agents integrate into the Lark ecosystem—not as floating windows, but as neurons embedded directly in business flows: reading production-scheduling data from multi-dimensional spreadsheets, retrieving standard operating procedures from documents, and becoming living components of operational continuity.
Only when AI ceases to exist as a standalone tool—and instead becomes woven into existing workflows—can collaborative work truly penetrate the heart of business operations.
Chapter 2: Letting AI Take Root in Authentic Business Flows
Over the past two years, as I’ve observed AI implementation on the front lines of business operations, I’ve noticed a widespread misalignment.
Everyone is racing to outdo one another in computing power and model sophistication—but in workshops and retail stores, what truly drives people mad is rarely an insufficiently “smart” model. Rather, it’s the lack of high-quality, context-rich business data. In many enterprises, real-world data remains scattered across siloed legacy systems. To get AI up and running, teams often need to manually construct artificial scenarios and fabricate datasets—ending up, after much effort, with little more than a visually impressive but functionally hollow “showroom.”
In contrast, within the Lark ecosystem, I’ve witnessed a markedly different path forward.
Lark provides foundational capabilities—including instant messaging, multi-dimensional spreadsheets, documents, calendar, video conferencing, and approvals—that users engage with daily and frequently. Over time, these tools organically capture the most authentic contours of business activity: communication context lives in messages, project evolution unfolds in documents, and operational workflows flow through multi-dimensional spreadsheets. Lark itself functions as a massive, living “business context container.”
When third-party AI assistants or intelligent work agents plug into the Lark ecosystem, they don’t need to go searching for data—they land directly in the richest soil available. On-site, I’ve seen employees invoke third-party AI assistants directly inside multi-dimensional spreadsheets to rapidly process complex production-scheduling data; or project managers casually enlist AI-powered office tools while drafting documents to clarify project logic.
No forced system integration—just seamless, intuitive business flow. Not “AI first, then hunt for use cases,” but AI plugging directly into already established collaboration scenarios. When AI grows naturally within every conversation and every line of recorded data, business intelligence finally gains tangible, grounded vitality.
Chapter 3: General-Purpose Large Models Are Still Premature—Embedding AI Into Existing Workflows Is the Faster Path
A few days ago, over tea with a manufacturing CIO, he posed a sharp question: “Is it still too early for AI to fully take over day-to-day work?”
Indeed, expecting a general-purpose large model to directly replace human employees remains, for now, a technological fantasy. But what if AI instead augments existing workflow touchpoints? That path isn’t premature—it’s precisely the right moment.
It’s analogous to the early days of “smart home appliances.” We didn’t begin by building an all-in-one domestic robot capable of handling every household chore. Instead, we gradually endowed air conditioners, refrigerators, and robotic vacuum cleaners with environmental awareness. The evolution of workplace tools follows the same logic. The future may not involve assigning each employee a superhuman virtual assistant—but rather equipping our current office tools with “brains.”
Within Lark’s open ecosystem, this “node-level intelligence” is already unfolding in practice. Third-party AI assistants and intelligent work agents are seamlessly integrating into daily routines—calendar scheduling, video meetings, approval workflows, and more. For instance, immediately after concluding a cross-departmental video meeting, users can effortlessly ask a third-party AI assistant to distill key action items; or, while processing an approval request, let an intelligent work agent verify critical compliance details.
No system overhaul—only natural extension atop existing workflows. While general-purpose large models continue searching for their ultimate form, embedding AI into current workflows has already become the fast lane for enterprise digitalization. Rather than chasing disruptive, all-at-once transformation, the focus is on delivering measurable efficiency gains at every granular business step—the most pragmatic stance for technology entering the real world.
Chapter 4: Infrastructure for Organizational Intelligence and an Open Ecosystem
In the past, we evaluated collaborative software primarily through the lens of “communication efficiency.” Today, however, it has fundamentally evolved into infrastructure for organizational intelligence.
Instant messaging, multi-dimensional spreadsheets, documents, calendar, video conferencing, and approvals—these seemingly ordinary, foundational capabilities that define Lark constitute the bedrock of enterprise digitalization. Yet delivering tools is always just step one; the real test lies in whether those tools can be absorbed into the organization’s very DNA.
I’ve spoken with numerous peers in physical industries, and one shared fear stands out: “system silos.” Production data from the shop floor, sales feedback from the front line, and delivery progress from the back end often reside in disconnected systems. Lark’s open ecosystem was explicitly designed to stitch together these fractures—supporting integrations with diverse third-party business systems and reweaving previously isolated processes into cohesive wholes.
When yield-rate data from the production line flows live into multi-dimensional spreadsheets; when customer feedback reaches responsible personnel via precise instant messaging alerts; when approval workflows connect and synchronize with external financial or ERP systems—the full closed loop—from business event occurrence, to data accumulation, to intelligent assistance—finally runs smoothly. Within this open ecosystem, third-party AI assistants and intelligent work agents find their natural foothold, becoming invisible yet indispensable helpers embedded in business flows.
True organizational intelligence isn’t about constructing a monolithic, hovering “super-brain.” It’s about enabling data and AI to circulate freely within every capillary of business operations.
Chapter 5: Collaborative Work Finally Reaches Deep Into Business Operations
That day, stepping out of the factory workshop, I glanced back at the electronic dashboard—its numbers flickering in real time: yield rates, production capacity.
Suddenly, it struck me: the organizational digitalization we’ve pursued so intently for years has always resided—in plain sight—within the roar of the production line.
Reflecting on the evolution of collaborative work, earlier tools centered on connecting “people”—making communication smoother and information more transparent. Later process systems shifted focus toward connecting “tasks”—standardizing approvals and ensuring compliant workflows. Today’s Lark ecosystem, however, truly connects “people, tasks, and intelligence.”
When foundational capabilities—such as instant messaging, documents, multi-dimensional spreadsheets, calendar, video conferencing, and approvals—cease to exist as isolated digital islands and instead seamlessly interface with third-party AI assistants and intelligent work agents, the very logic of organizational operation transforms. People no longer need to awkwardly adapt to rigid systems; instead, intelligent systems adapt to assist people.
I’ve discussed this with many peers. Many once worried that AI’s arrival would intensify workplace “technological anxiety,” even reducing people to gears driven by algorithms. In fact, the opposite is true. The ultimate goal of collaborative work has never been to manufacture anxiety through flashy tech—but to achieve deep integration between the digital world and authentic business contexts. Let data fill the gaps; let intelligence handle the legwork; and let people reclaim their highest-value role: creative decision-making and insightful judgment.
From connecting communication, to connecting business, to connecting intelligence—the shift this time may transcend tools themselves. Collaborative work, at last, is truly reaching deep into the core of business operations.