A thousand times asked, how AI agents and AI operations are reshaping business models

This article aims to start the discussion, talking about the subtle clues of AI large models in the consumer end and business models. Alibaba's large model, the rise of Qianwen's AI Assistant and AI Agent, gives me some insights: 1. In the past, systems were for humans to use; in the future, they will primarily be for silicon-based AI to call upon; 2. Consequently, product forms need to be completely redesigned; 3. Brand new consumer demands are quietly gestating.

—— Introduction

—————

At this very moment, light flickers while consumption wanes. A few phrases everyone knows are: You should stand in the light, don't just stand there, and definitely don't stand there naked; but my friend—that bird guy who developed hyperlipidemia from eating at Shandi—says: All that is correct, but the most, most, most important thing is, don't stand naked in the light.

After all, he's a winner in life, suddenly terrifyingly rational, and I can only approve. No matter how optical communication embraces the peak of its cycle, large models ultimately must land in consumption, otherwise your closed loop can't bring in that final payer. Your consumption is my income, without consumption, where does income come from? All future free cash flows scheduled fifty years ahead are just illusory bubbles.

And large models + mass consumption, AI Agent driving consumption, this area is the strength of Chinese concept stocks, Alibaba's strength, and the business logic is also the most straightforward, basically amounting to, pulling out the truth and pulling the trigger.

I'll continue using language even my grandma can understand to sort out the product logic and business logic within.

I saw Qianwen integrated with China Eastern Airlines, so from flight inquiries, to ticket purchase, seat selection, check-in, to refunds and changes, Qianwen's AI Assistant can provide a full-chain service—all you need to do is shoot out natural language (i.e., speak human). Previously, Qianwen integrated with food delivery, ride-hailing, movie tickets, hotels, Taobao flash sales, Gaode, Fliggy, etc., all were internal verifications within the Alibaba ecosystem; integrating with China Eastern Airlines is the first external output of its AI Assistant capability. We can reasonably expect that Alibaba's AI Agent external large-scale deployment has begun.

Looking at this from a business model perspective: Qianwen's first-time opening of AI Assistant capability indicates that any APP, any product, any service capability can be Agent-ified. Then the entire product model becomes: driven by Homo sapiens using natural language to operate AI, and then AI using non-natural language to mobilize the entire data system (for example, to serve a complex personalized need covering multiple parameters like budget, seat selection, date, aircraft model, etc., requires AI to call upon and combine multiple systems to complete the task and do it better than a human).

This also hints at another trend: all products and services in cyberspace will be remade to make them easier for AI to use. Because the shaping of previous product forms was entirely for the convenience of Homo sapiens, and now we need forms convenient for AI use. I'll tentatively call it: we need to establish a tokenizer for human actions, a semantic index of business logic. Why do we need a tokenizer?—Because the form of information humans can process is different from what machines can process, just like large language models can only process tokens, but cannot directly handle natural language.

To put it more plainly, the design and shaping of systems in the past were for human use, while future systems are for AI use. And between all the intricate, entangled systems and individual specific users, the AI large model plays the role of a tokenizer that links, slices, maps, translates, and decodes. Isn't this exactly the traffic portal that the masses loved in the last web X.0 era?

So, in essence, it's a reconstruction of traffic portals. Whoever can most quickly complete "using/for AI to remake all products and services" will be the master of the next-generation traffic portal. And the previous mobile internet business model will undergo a disruptive upheaval here. But the prerequisite is that you have the capability to completely reshape the product forms. Or you can also, like Claude, use computer use to completely take over your terminal, completely operate your computer or phone, but this is just like 100% autonomous driving—many people still can't get past that psychological barrier.

Better to proceed step by step, let AI do some small things, some practical things first. And Alibaba's current advantage lies in its own massive closed-loop ecosystem.

Furthermore, AI Assistants like this will bring new changes on the demand side, because natural language will stimulate new demands and growth.

Human natural language has many descriptive parts, like adjectives—azure blue, tranquil, bizarre—they were naturally filtered out by systems because they severely increase system complexity. When you used to open an APP to buy a plane ticket and select a seat, you could only see a crude schematic of airplane seats and seat numbers; you didn't know which seat offered a view of the starry sea or the changing seasons. But now you can tell the AI:

I'm traveling with two kids whose urination schedules are utterly bizarre, and I want a few quiet, non-noisy seats, preferably with a view of the azure blue sea.—Thus, Qianwen might arrange for you to sit in a triple seat near the tail of the plane, with a restroom just inches away, no matter how unpredictable your bathroom schedule becomes.

These demands, which were once hidden to comply with machine language requirements—these diverse, overwhelming, richly enchanting, varied, and colorful personalized demands—will be further stimulated by the revolutionary assistance of AI agents. This is where some inherently highly standardized business forms might find differentiation and growth curves.

After all, sluggish consumption truly needs fresh stimulation. Just a few thoughts in this article, not necessarily correct.