On July 30, ByteDance did something that would have been unthinkable a year ago. It gutted Feishu—its proud, decade-old enterprise collaboration platform—and stuffed the pieces into its AI chatbot, Doubao. The product team was absorbed. The sales team was handed to Volcano Engine. Feishu, once a crown jewel alongside Douyin, was effectively dismantled.
If you’re an enterprise buyer staring at a six-figure AI office contract, this should terrify you. Because the company that spent billions on AI infrastructure just admitted that the standalone office app is dead—and that AI office, as a product category, is a mirage.
You’ve probably noticed the pattern. Your team uses AI to draft documents faster, but the final version still takes the same amount of time. The AI generates a report, then you spend twice as long fixing its hallucinations. The promised productivity gain? It’s just shifted the work from creation to verification. Total labor time: unchanged.
Let’s call it what it is: the efficiency illusion. And it’s the reason ByteDance, Tencent, and Alibaba are all frantically reorganizing their AI office products—not because they’ve found a winning formula, but because they’ve run out of alternatives.
The $40 Billion Trap
Here’s the math that keeps AI executives up at night. Doubao has 200 million monthly active users. It generates less than $1 million in daily revenue. Its daily compute cost? Tens of millions of dollars. The more users Doubao gets, the more money it loses. In traditional internet logic, users are assets. In AI logic, users have become a liability.
ByteDance’s 2026 capital expenditure is over $200 billion—roughly 60% of its 2025 profit. Every dollar of AI revenue comes with a dollar of compute cost, and usually more. The C-end (consumer) market can’t sustain this. So the narrative shifts to B-end: sell AI to enterprises.
But enterprises don’t buy AI—they buy results. And the results from current AI office tools are deeply unreliable. A financial report with a hallucinated number, a contract summary that misses a critical clause, code that introduces a security vulnerability—the cost of mistakes in enterprise environments is far higher than the time saved. 99% accuracy is 100% unusable in core business processes.
Why the Giants Are Merging, Not Winning
Look at the three big moves. Tencent consolidated its three agent products into WorkBuddy. Alibaba merged its three into Qianwen Office. ByteDance dismembered Feishu and fed it to Doubao. All three are doing the same thing: admitting that a standalone AI office product has no future.
But here’s the twist. The real value isn’t the AI—it’s the data. Feishu spent ten years building the organizational context of hundreds of enterprises: permission structures, workflows, historical documents, chain of command. That data is the only thing that can make an AI useful in a business setting. Feishu wasn’t a failure. It was a ten-year investment in the fuel that AI needs.
ByteDance’s strategy now is to sell Doubao’s enterprise model with Feishu bundled as a free add-on. Customers don’t want to buy Feishu; they want AI. So ByteDance gives them the AI, and the office tool becomes invisible. That’s the real future: AI that disappears into the workflow, not an app you have to consciously open.
The Practical Bottom Line
If you’re a decision-maker evaluating AI office tools, stop asking ‘Can this tool save me time?’ Start asking ‘Does this tool actually reduce my headcount or just create more busywork?’ The evidence so far suggests the latter. Companies that adopt AI office tools often end up with the same number of employees, just with more drafts to review.
And if you’re a product manager, understand the strategic trade-offs behind ByteDance, Tencent, and Alibaba’s moves. They’re not building for the product you see today. They’re building for a future where the AI is embedded so deeply that you don’t even notice it. The ‘super app’ for AI office doesn’t exist—and it never will. The real revolution is when AI becomes so natural that we stop talking about it.
ByteDance’s decision to dismantle Feishu looks like a loss. But it’s actually a recognition that the old game is over. The new game isn’t about building the best office software. It’s about making AI so invisible that the office itself fades away. And if you’re still buying a ‘next-gen AI office suite,’ you’re paying for a ghost.
FAQ
Q: But isn't there at least some productivity gain from AI office tools? I've seen my team write emails faster.
A: Yes, but only for low-stakes tasks like email drafts or meeting notes. The moment you enter core business processes—contracts, financial reports, code—the accuracy drops and the review time explodes. The net gain is near zero. The real cost is the hidden overhead of verification.
Q: If AI office is a pseudo-proposition, why are all the giants investing billions? Just to fool investors?
A: Partly, yes. They need a B-end narrative because C-end AI is bleeding cash. But there's also a long-term bet: that models will eventually become reliable enough to eliminate human review. However, that day is not here, and the window for profit is uncertain. Most are buying optionality, not a viable business.
Q: So what should I do instead of buying an AI office suite? Just ignore AI?
A: No. Use AI for specific, low-risk tasks. But don't replace your workforce with it. Focus on building your organizational data infrastructure—clean workflows, good permissions, structured data. That's what will make AI useful when it finally works. For now, the best investment is in data hygiene, not AI tools.