Over the past two years, nearly all enterprises have vociferously championed AI transformation. Yet, most implementations have been relegated to employees' personal "writing aids" or "customer service pop-ups." Where, precisely, do these grand AI initiatives, costing millions of dollars, falter?

A few days ago, a friend of mine who operates an e-commerce agency lost a potential million-dollar client they had been courting for most of the year.

The client was negotiating a Double 11 (Singles' Day shopping festival) collaboration. When it came time to finalize the deposit and secure warehouse space, the bill exceeded expectations by nearly 100,000 yuan. The client questioned why the previously agreed-upon "tiered rebates" had not been applied; the draft sales contract specified "applies only to main bestsellers"; the warehouse had stocked up based on "buy-one-get-one combo packages"; and the newly hired accountant, unaware of any pre-sale rebates, was merely applying formulas from the previous accountant's Excel sheet.

Sales, warehouse, and finance each presented evidence from their own systems to absolve themselves of blame. However, after assessing the situation, the client deemed the company's internal management chaotic and feared potential issues during Double 11, leading them to decisively terminate the collaboration.

The most taxing aspect of the workplace often arises from this type of "contextual misalignment."

Many enterprises believe that acquiring an AI tool can resolve all efficiency issues. However, over the past two years, numerous companies have vociferously advocated for AI transformation, only to discover, as in the case of my friend mentioned earlier, that despite investing in various AI assistants, these tools ultimately devolve into personal writing aids or customer service pop-ups, failing to bolster core business operations.

Where exactly do enterprises falter when implementing AI?

At yesterday's (September 22) Cloud Town Conference, Chen Yusen, CEO of Qwen Office, offered a keenly insightful perspective:

Just because personal office tasks become more streamlined doesn't equate to improved organizational efficiency. After three years of large language model development, the primary reason they haven't been integrated into business workflows is the absence of "Enterprise Context."

Qwen Office's solution is straightforward: abandon the pursuit of an "omniscient" conversational assistant and instead transform AI into an imperceptible conduit that links fragmented enterprise contexts scattered across group chats, documents, approvals, and even verbal communications.

This is the rationale behind Qwen Office's approach—Context is All You Need. The entire presentation unveiled the future form of enterprise agents.

01. Why Can't AI Penetrate Core Businesses? The "Transition" Falters

Enterprise business AI and personal AI are fundamentally different.

Utilizing AI personally is straightforward—chat, close the window, and it's done, with no lasting repercussions. However, enterprise operations are ongoing, involving interconnected steps from receiving requests, checking inventory, calculating pricing, to processing approvals.

Previously, AI was confined to translation and writing tasks, struggling when confronted with real business processes due to fragmented cross-departmental and cross-system contexts.

Take Yihai Kerry Arawana, for instance, which manages a diverse portfolio of brands like Arawana and Huiji Flower. Previously, creating a dish proposal for Catering clients required cross-system coordination among multiple departments, involving over 300 steps and taking several days. For financial specialists to calculate new product pricing, they had to manually gather data on formulas, market trends, inventory costs, and shipping fees for repeated calculations.

Why wasn't AI utilized before? Because it couldn't access these core enterprise data and processes. Qwen Office addresses this by employing open-source MyContext to package scattered data from documents, approvals, and messages into an "organizational context" that agents can access anytime.

Now, when creating dish proposals, they input the client brand and development stage to receive reports within hours. For pricing calculations, submitting a standard form in Qwen Office yields results instantly.

Similar scenarios unfold in the logistics industry. Previously, China Data Link's staff had to toggle between highway, shipping, aviation, and port platforms to track freight progress or conduct foreign trade analysis, with information fragmented across systems.

Now, by inputting a single request in Qwen Office, they can leverage logistics industry-specific capabilities to track shipment progress, query port clearance status, and analyze foreign trade trends.

Only by linking cross-system contexts can business "transitions" proceed smoothly, enabling AI to evolve from a mere chat companion into an automated conduit that completes business processes.

02. Why Must the Premier Enterprise AI Be "Invisible"?

Since context is essential, how is it acquired?

Previously, many enterprises made the common error of "revolutionizing" their operations during informatization or AI transformations—purchasing costly new systems and compelling employees to abandon long-established habits by adopting new forms and software.

Countless painful experiences attest that this approach nearly always fails due to the strong inertia of entrenched organizations.

Qwen Office's most astute move is avoiding confrontation with existing systems by employing "invisible management."

First, it doesn't create new entry points but infiltrates existing platforms.

It doesn't require employees to download new software; instead, it integrates capabilities directly into mainstream platforms like DingTalk, Feishu, and WeCom, even connecting with Salesforce on Alibaba Cloud. Agents operate wherever employees type and manage CRMs, eliminating additional adaptation costs.

Second, it transforms verbal communications into business instructions without altering old habits.

Much of an enterprise's core decision-making and business details reside not in systems but in verbal exchanges during meetings, client visits, or hallway conversations—the "vanishing context."

Qwen Office released a 67-gram portable card, QwenNote A2. For example, partners at law firms like Fangda or DeHeng, after meetings or court sessions with clients, don't need to spend hours at their computers organizing notes. They simply press and hold the card and say, "Organize the recent communication into a case memo, highlight three pieces of evidence needing supplementation, and schedule a team meeting this Friday."

The AI, having captured the conversation context, immediately collaborates with DingTalk to organize the memo, create a checklist, and schedule the meeting.

This approach doesn't alter how people work but enhances their efficiency.

Third, it ensures seamless security and compliance without leaving traces.

Enterprises hesitate to feed context to AI due to fears of data leaks. Financial institutions like Guoyin Financial Leasing or law firms like Fangda have stringent data compliance requirements.

QwenNote employs "burn-after-transfer" at the hardware level—encrypting and transmitting audio to the cloud for transcription, then physically deleting the original audio while retaining only the text summary. Combined with role-based permission controls at the platform level, enterprises can confidently entrust business contexts to AI.

03. With Context, AI Implementation Becomes "Lightweight" and "Deeply Rooted"

The most practical implementations are often exceedingly lightweight.

Once contexts are linked, enterprises don't require multi-million-dollar "grand transformation plans" to utilize AI. The most effective path is to effortlessly compile employees' personal "workarounds" and experiences into reusable corporate "hard assets."

Consider these real-world examples:

At Changan Automobile's development center, electrical system integration test engineer Mao Hai previously spent two days manually calculating wiring harness selections. Using Qwen Office, he built a tool based on his business experience, reducing the task to just five minutes. Purchasing department employee Gong Xuan, who used to work late nights preparing weekly meeting PPTs, now delegates the task to AI, securely retaining sensitive data locally.

At FAW Toyota, frontline workshop team leaders without programming backgrounds use Qwen Office to verbally describe production management systems covering multiple lines. Sales managers automate dealership daily reports, inspections, and order deliveries, compressing data processing from a full day to minutes.

At Fangda Law Firm, previously, dozens of lawyers spent weeks manually verifying corporate legal entities across hundreds of data sources before IPOs or acquisitions. Now, partners solidify review frameworks into Skills, enabling agents to automatically batch-access websites, verify data, and generate reports with just one lawyer's final review.

In none of these instances did major overhauls or compelling employees to code occur. Essentially, employees' mental expertise and workplace workarounds were transformed into reusable digital assets with minimal barriers.

Chen Yusen stated at the Cloud Town Conference: "In the PC era, the work language was files; in the mobile internet era, it was online collaboration. In the Agent era, the new organizational language is 'enterprise context.'"

From underlying cloud computing power with tens of thousands of cards, to MyContext for data integration, to frontend hardware and multi-user workstations, Qwen Office aims to construct a "digital nervous system" serving organizations.

When personal expertise seamlessly transitions into organizational assets and isolated efficiency gains connect into business workflows, enterprises' evolution toward AI-native operations naturally ensues.