AI Office Tools Are a Garbage Race — So Why Are Big Tech Giants Fighting Over It?
Lately, every elevator is blasting brainwashing ads for some "Work" or "Buddy" tool. Qwen Office, TraeWork, WorkBuddy—these AI office tools are duking it out in a bloody three-way battle.
But let me drop a hot take right off the bat: competition in AI office tools is basically meaningless. This is a total "garbage race"! (At least in China.)
So if it's a garbage race, why are Alibaba, Tencent, ByteDance, and even Google in Silicon Valley all throwing money at it? Is it really just human depravity?
If you see this as just an efficiency tool for office workers, you haven't even touched the surface of this game.
Stop Pretending—You're All Here to Burn Tokens
One thing is absolutely clear: once you bring AI into the office, especially into work group chats, your token consumption explodes exponentially compared to before.
Before, AI was one-on-one: you ask, it answers. Now AI is in the group, and what does it do? It has to read the context! Whether it does any work or not, it has to chew through hundreds of rambling chat messages and messy meeting notes, digest them, and spit out deliverables. That consumption is tens or even hundreds of times more than you alone using Doubao to watch fishing!
The cloud business units at big tech are probably laughing in their sleep at this skyrocketing consumption: finally, some suckers to burn tokens by the truckload!
Remember how C-end used to be? Everyone was used to freeloading and refused to pay. Now, C-end is diving into the office market, fighting for entry points and users. On the surface, they're offering free AI assistants. In reality? They're burning way more tokens. This is the real "wool from a pig, paid for by a dog"—the cloud business is raking it in!
Say Goodbye to Pig Questions—Big Tech Wants Your "Work Data"
Besides burning tokens, big tech is also after your body—specifically, your high-quality process data.
Let's look at the real C-end situation. You think the top models trained with billions of dollars and countless A100 GPUs are being used for what? I'll tell you: asking "how to raise pigs" and "help me watch the fishing line"!
What do you take this for? Big tech bosses would vomit blood if they saw the backend data: I invested all this money to train models, just to solve your boring problems? What the hell!
But the office scene is completely different. Every conversation, every task assignment, every weekly report summary is extremely precise, with clear business goals—this is "high-quality work data." You can't buy this kind of data on the open market!
Now, just offer a free office AI, and users will willingly feed it the company's core operational logic and communication context. Using this data for post-training is priceless. You think you're freeloading on a high-efficiency tool? You're actually working as a free "data annotation laborer" for big tech—and bringing your own lunch! You want to run? No way!
Don't Argue—Silicon Valley Already Proved This Play
You think this is just Chinese big tech going crazy? Dead wrong! Silicon Valley has already slapped everyone in the face with real data.
We have to look back at Anthropic's Claude Tag. This is definitely the bellwether for the entire AI industry this year, and it's a god-tier move that China has severely underestimated. Even Karpathy came out and said this is a whole new generation of AI interaction revolution.
Why is it so amazing? Because Anthropic's backend data shows that 60-70% of product iterations and code PRs are now completed directly in Claude Tag's group chat environment! People don't need to open a separate Claude desktop app or type in a Terminal command line.
In enterprise chat software like Slack, AI is like an omnipresent ghost. Just @ it, and it can instantly get the job done based on the entire channel's chat history and context. This is why I say "context" has completely crushed "pure intelligence." A few years ago, we thought AI wasn't smart enough and kept failing, so we chased higher parameter counts. But now? The underlying models at all big tech have long passed the "good enough" threshold. In real office scenarios, an AI that can read human bickering in a group chat is ten thousand times more useful than a nerd who can only solve Olympiad math problems!
The "Cyber Overseer" at the Dragon Gate Inn
This also triggers another terrifying reality: office AI has now openly become the innkeeper of the "Dragon Gate Inn" in group chats!
What does that mean? Before, AI was a temporary tenant; we could only give it a few rooms. Now, AI is an inn with an open kitchen. What projects your company advanced today, who killed whom, who's been "read but not replied" in the group—the innkeeper sees it all! It can not only capture all fragmented information but also sort out clear swimlane diagrams.
What's the scariest part? It can even hold a weekly meeting and call you out for a flogging: "Why didn't you do these three tasks? Explain yourself!"
Before, we sent messages in group chats to provide emotional value to the boss, proving "I'm working hard and responding." Now? Everything we chat about is for AI to see! This is a hellish cyber overseer! You want to resist? Dream on!
The "Cloud Abacus" Behind the Three-Way Battle
Look at the current chaos: Qwen Office, TraeWork, WorkBuddy. Are they really competing on features?
Take Qwen Office: it just hands you a "bucket-type" Qwen-Max and doesn't even let you choose the model. Why? Because backed by Alibaba Cloud and DingTalk's ToB foundation, it knows exactly what enterprises want. In general office work, "good enough" is all you need; no need for you to pick and choose.
Then there's WorkBuddy. At first, I couldn't understand why it got popular. But then I got it: it has a large free tier and lets you switch models freely. For those penny-pinching side-hustlers, it's a godsend. Even though it's backed by WeChat's AI-unfriendly ecosystem, it still grabbed a huge chunk of lower-tier users with "good enough" and "first-mover advantage."
TraeWork is even more obvious: deeply integrated with Feishu. If you have AI send a report, it's tagged with "from TraeWork." Your account is AI, AI is you—this is the true cyber avatar.
But behind all this, who's the real winner? It's all the cloud business units!
If you dig into the equity structure on Tianyancha, you'll find a brutal truth: Qwen Office is Alibaba Cloud's; WorkBuddy is Tencent Cloud's; ByteDance's Doubao commercialization all goes to Volcano Engine. All AI product departments now have one job: go generate traffic and create scenarios! As for collecting money? That's the cloud business's job!
AGI Dreams Shattered—Monetizing Compute Is the Endgame
This isn't just Chinese big tech's play; Silicon Valley has secretly swapped scripts too.
You think Google's recent cloud business surge of over 50% is due to what? Not some earth-shattering AGI breakthrough, but aggressively pushing small models like Flash to eat up everyday scenarios! Google is probably furious with its research scientists: "Hassabis, Jeff Dean, stop messing around with our research dreams! I'm here to make money! A top model that's been in development for months? Forget it, I just need to spread entry points and consume compute!"
Look at Musk's Grok, which just launched Grok Bot in office scenarios. Why? Because they bought too many cards! "Originally there was nothing, where does dust come from?" Since we can't catch up to the top intelligence, let's just sell compute to enterprise office! Isn't Meta the same? They open-sourced their closed model after all that effort—why? Because they found open-sourcing sells compute faster!
So, while we mock AI office as a hyper-competitive garbage race, big tech is quietly completing the perfect business loop behind the scenes.
Imagine if, at some point in the future, the cost of AI completing a specific office task could be precisely priced—like DeepSeek running benchmarks for just $0.03. That would be huge! The boundaries would become crystal clear: all the clearly-priced work goes to AI, and the rest goes to those salaried folks handling human relations. By then, this massive cloud empire will completely take over all global enterprise IT spending.
Of course, for us ordinary workers, watching our daily group chat bickering turn into skyrocketing token consumption and cloud revenue in big tech's earnings reports feels like... being harvested while you're down, and then having to reply in the group with perfect business etiquette: "Received, thanks boss!"
At the end of every AI myth, there's the compute cash register!
[Subscribe to JustSayAI Daily Brief] 📮JustSayAI Daily Brief · Fan Exclusive Benefits Two emails daily · Listen in one click! Now supports Email + IM + Skill endpoints