Graphic | Sister Tang

The AI Agent craze, initially sparked by OpenClaw at the beginning of the year, has now surpassed the six-month milestone. During that time, discussions about Agents were ubiquitous across the Chinese internet. Major companies rolled out OpenClaw-like products, while startups flocked to develop desktop assistants, as if everyone believed they could secure a foothold in this emerging market.

Looking back six months later, Tencent's WorkBuddy has truly capitalized on this demand, establishing itself as a prominent personal office entry point.

What sets WorkBuddy apart is its approach; it didn't rely on a globally leading proprietary model to stand out. While Codex and Claude Code leveraged top-tier foundational models, WorkBuddy opted to integrate multiple models, Skills, local tools, and Tencent's ecosystem at the product level. For those without programming experience, it's a product well worth exploring.

Although I'm not a WorkBuddy user myself, I can relate on a personal level. When I first used Claude's desktop version earlier this year, I never touched Claude Code because I didn't think I needed programming. However, after actually using it, I realized it could handle files, tools, and multi-step tasks within my existing workflow, making it useful for both research and daily office work.

WorkBuddy's success on the personal front at least confirms one thing: beyond competing on models, delivering a great product is also a viable strategy. Around the same time, Alibaba and ByteDance began consolidating their Office Agent resources. On August 3, Alibaba integrated three Agents into Qianwen Office and launched a public beta; on August 6, ByteDance emphasized AI productivity and enterprise business at its mid-year all-hands meeting, having already reorganized teams related to Doubao, Feishu, and Volcano Engine.

However, the excitement on the personal front is still far from translating into a business that generates sustainable cash flow. What's most visible now are features, free quotas, and social media presence, while payment scale (paid scale), usage depth, renewals, and unit costs remain difficult to assess.

Moreover, Office Agents face a natural disconnect: employees use them, but the work results belong to the company. For a single report, an individual might occasionally purchase some Tokens; but when Agents are used long-term, access internal data, and act on behalf of employees, costs, permissions, and responsibilities should no longer be left to the employee. To truly scale, they must eventually fit into corporate budgets.

01 Usability Is Just the Starting Point

The reason Office Agents currently seem like a personal competition isn't complicated. Employees are the first to discover new tools and are easily swayed by Skills, multi-model support, and free quotas. Clicking a Skill to watch it write a report or create a spreadsheet delivers value within minutes.

Such demonstrations are crucial—they transform an unfamiliar technology from something incomprehensible to something tangible and user-friendly. But personal ease of use only proves the product has cleared the usability barrier; it doesn't mean the entire workflow is ready for enterprise production, let alone that enterprises will pay for it.

What individuals feel most acutely are immediate, visible results. Having an Agent write reports, organize spreadsheets, or handle code is something today's foundational models can already accomplish to a significant extent. But what enterprises pay for isn't these isolated results—it's whether the Agent can handle a specific job end-to-end.

The same customer, order, or profit metric might use different definitions across departments within the same company, with knowledge scattered across customer systems, business systems, data lakes, and legacy systems. While models can write a coherent report using public data, inside an enterprise, they must adhere to the company's definitions and deliver usable results.

Even with correct data and definitions, the job is only half done. Reports need approvals, data must be written back to systems, tasks may stall and require human intervention, and work must resume from the right point afterward. The Agent isn't just solving a problem—it's handling a process that can be interrupted, fail, and must keep moving forward.

As workflows lengthen, costs rise. Amazon Web Services (AWS) example calculations show that when a single request requires three tool calls, inference costs can reach 5–10 times that of a single tool call, as each round sends accumulated conversation and tool results back to the model. Additional orchestration, storage, network, and security loads also influence how long enterprises are willing to let Agents run and how many tasks they handle.

If Agents can spend money, modify systems, or take external actions, enterprises face more than just costs. Who grants permissions? Who checks results? Who takes over if something goes wrong? These must be clarified before action. An individual can redo a task after an error, but a company must ensure a single mistake doesn't propagate through subsequent work.

Costs escalate with task complexity, and responsibilities expand with actions. Naturally, enterprises won't initially hand over entire workflows to Agents. A more realistic and common approach is to first define an observable, controllable scope, then gradually expand AI tool usage and authorization based on results.

Banco Bilbao Vizcaya Argentaria (BBVA) began collaborating with OpenAI in May 2024, initially granting ChatGPT Enterprise access to about 3,300 people, expanding to around 11,000 by May 2025, and announcing a rollout to about 120,000 seven months later. By June this year, actual usage exceeded 100,000, two years after the first accounts were activated.

JPMorgan Chase moved faster. It launched LLM Suite in summer 2024, placing large models in a secure internal environment, covering about 200,000 people from zero in eight months. But this first established a unified, secure employee access point for models, not that the enterprise had handed over data, systems, and action permissions to Agents all at once. Deeper data integration and workflow authorizations would still require future steps.

While the entry point is managed by the enterprise, the Agent's operational resources may span both internal and external environments. Enterprises must decide which capabilities can leverage external resources and which must remain internal. IDC and the China Academy of Information and Communications Technology surveyed 250 enterprises, finding that among IDC's surveyed intelligent computing cloud service users, 85.8% adopted a hybrid architecture combining external AI computing power with their own data centers.

These choices, while varied, follow the same logic: the organization must regain control over what Agents can access and how far they can go. Models provide capabilities, but a product mechanism outside the model is needed to organize tasks, verify results, and restrict or take over actions when necessary.

02 Who Can Turn Implementation Experience Into a Product

Even with the same model, different products can achieve vastly different outcomes. Some products require constant human reminders at every step, while others remember progress, call tools, and stop when errors occur. These differences aren't encoded in model weights but determine whether Agents can deliver results.

The product runtime framework (Harness) organizes model capabilities into deliverable products. It feeds task-required materials to the model, arranges steps, calls tools, remembers progress, and checks and retries when results are incorrect. When AI Agents enter enterprises, identity, permissions, and auditing must also fit into the same mechanism.

The stronger the model, the less the Harness needs to replicate what the model already does well. Product design should instead focus on verification, recovery, and governance.

In July 2026, Anthropic revealed that for Claude 5, it adjusted Claude Code by removing over 80% of system prompts without performance loss in programming evaluations, then added necessary mechanisms where the model repeatedly failed. Boris Cherny, head of Claude Code, also argued that enabling models to verify their own work during execution is one of the most underrated capabilities today.

Unlike personal users, when Agents enter enterprises, the product layer beyond the model extends from a single product to a cloud platform. Cloud vendors combine model services, runtime environments, and organizational governance into one layer. In April 2026, Google Cloud announced that Vertex AI would be delivered via the Gemini Enterprise Agent Platform.

Breaking it down: a single Agent's task completion is organized by the Harness; managing multiple Agents across an enterprise requires a unified registration, authorization, and management system—the enterprise Control Plane. Domestically, Tencent Cloud's WorkBuddy Managed Agents, Alibaba Cloud's Bai Lian and AI Gateway, and ByteDance's Volcano Engine AgentKit and HiAgent are also building similar organizational governance capabilities through different product combinations.

While Harness and enterprise Control Planes can enforce rules, they can't create them for the enterprise. When Agents are used within a single team, definitions, permissions, and acceptance criteria can be locally agreed upon; but once they cross departmental boundaries, these rules must be realigned organization-wide.

Anthropic's survey of over 500 U.S. enterprise technology leaders quantified this divide: 57% of respondent enterprises already used Agents for multi-stage processes, but only 16% had entered cross-team workflows. In the same survey, 46% cited existing system integration as a major obstacle, while 42% mentioned data access and quality.

Crossing this organizational boundary requires both system integration and data alignment. A 2026 KPMG report stated that for enterprise data to be usable by Agents, it must be searchable, carry sufficient business context, and establish a trustworthy foundation through data lineage, machine-readable permissions, and decision rules. However, data meanings, quality standards, and permission ownership are controlled by different departments.

For Office Agents to enter enterprise production, two transitions must occur. The Harness transforms the model into a product actually usable by the enterprise, while the enterprise embeds the product into its production systems and workflows. The first determines whether the Agent can complete tasks; the second determines whether it can enter formal production and lead to sustained procurement. Agent vendors can provide connectors, permission tools, and evaluation methods, but data interpretation, business definitions, and result acceptance must be consensus-driven by the client enterprise.

At this stage, each enterprise's unique data, permissions, workflows, and acceptance criteria first increase the implementation burden for individual clients. Switching to a new client requires vendors to reunderstand the business, reconfigure systems, and reverify results, with only partial reuse of experience from previous clients.

Compared to traditional SaaS, Office Agents cover broader work scopes and theoretically larger market boundaries. Currently, different vendors have arrived at this stage with different strengths. Cloud vendors and enterprise software companies are closer to data, systems, and identity frameworks, while independent Agent products can rely on cross-model, cross-cloud neutrality or deep industry expertise. Multi-model support alone isn't a moat; only by managing quality, cost, permissions, and evaluation together can model routing deliver enterprise value.

Tencent, Alibaba, and ByteDance's advantages in personal traffic, internal promotion, and cloud resources are undeniable. But a replicable business ultimately requires validation from third-party clients. Only when enterprises outside their ecosystems are willing to allocate budgets, data, and action permissions does this competition truly enter the enterprise market.

03 Conclusion

Popularity on the personal front generates buzz, but only the enterprise side leaves behind quantifiable, pricing-ready data.

For investors, the rise of Agents isn't entirely positive—it's turning the already opaque cloud and software businesses into an even deeper black box. Previously, it was hard to know whether revenue came from computing, storage, databases, or software; now, application seats, model inference, enterprise Control Planes, and implementation delivery are lumped into a single "AI business" category by many companies.

Microsoft disclosed an AI annualized revenue run rate exceeding $37 billion and over 30 million paid seats for Microsoft 365 Copilot; Salesforce disclosed Agentforce's annual recurring revenue (ARR) and transaction volumes, while Chinese vendors publicly share more about active rankings, seat prices, and the proportion of Feishu new customers purchasing AI products.

All seem to prove AI is generating revenue, but these metrics haven't become more granular alongside the numbers. The market still can't distinguish whether what's being sold is a replicable product or a project requiring repeated implementation.

Investors must answer two key questions: Has the Agent moved beyond seats and pilots into formal client production? Has enterprise deployment begun to achieve economies of scale? The former can be measured by production workflow proportion, manual takeover rates, and renewal expansion; the latter by how much implementation and customization resources each client consumes, as well as the onboarding speed and template reuse rates of subsequent clients.

More granular disclosures won't automatically boost stock prices, but they serve as operational signals. Companies willing to disclose production usage, delivery reuse, and client expansion at least make this business comparable; those only willing to share seats, Tokens, and contract numbers must accept market discounts for the opaque portions.

Disclaimer: This article is for learning and communication purposes only and does not constitute investment advice.

Like, share, and repost—your support keeps us updating!