On July 10, OpenAI dropped two big things at once: the GPT-5.6 model family, and ChatGPT Work.

Most people's attention gravitated toward GPT-5.6 — three tiers (Sol/Terra/Luna), Ultra multi-agent mode, capability improvements here and there. But in my view, the actually important announcement is ChatGPT Work.

Because it marks a strategic pivot for OpenAI: from "conversational AI" to "work AI". The old ChatGPT was you-ask-it-answers, one question one response. ChatGPT Work is you give it a task, and it plans on its own, executes on its own, gets it done on its own — working continuously like a real employee.

This isn't a difference in degree. It's a paradigm shift.

What Actually Is ChatGPT Work?

The official definition is "enterprise-grade agent," but that's too abstract. In plain English: ChatGPT Work is an AI employee that can run for extended periods, handle complex multi-step tasks, and make autonomous decisions.

A few examples of what it can do: you tell it "analyze our sales data from the past three months, find the reasons for the decline, and make a PowerPoint report" — it pulls data from the database on its own, analyzes on its own, finds root causes on its own, builds the deck on its own, and hands you the finished product. No step-by-step guidance needed from you.

Another example: you tell it "go through all this customer's emails, extract key information, generate a customer profile, then write a reply email" — it digs through emails on its own, summarizes on its own, analyzes on its own, writes the email on its own. You just take a final look and hit send.

What's the difference from old ChatGPT? Old ChatGPT was an "assistant" — you say one step, it does one step, you have to watch the whole time, correct it when it's wrong, pull it back when it goes off track. ChatGPT Work is a "colleague" — you hand it a task, it figures out how to get it done. It might come to you with questions if it hits issues, but most of the time it handles things itself.

Powering this capability is GPT-5.6 Ultra's multi-agent mode. Simply put: when a task comes in, it's not one big model doing everything from start to finish. The system automatically breaks down the task, calls multiple specialized agents, they collaborate by division of labor, and finally aggregate results. Just like a team — some do analysis, some write code, some do design, some write copy — everyone works together to get the job done.

A New Battleground in the Enterprise Market

The launch of ChatGPT Work means OpenAI is officially bringing the fight to the enterprise market.

What did OpenAI's revenue mainly come from before? Individual subscriptions (ChatGPT Plus) and API calls. Individual subscriptions have large user bases but low ARPU. API revenue is decent, but customer retention is unstable — today your model is strongest and everyone uses you; tomorrow someone cheaper comes along and everyone leaves.

The enterprise market is different. Once a company builds its workflows on your platform, connects its data, trains its employees — the migration cost becomes extremely high. It's a market with incredibly strong stickiness, extremely high ARPU, and extremely high lifetime value. Before, this market was Microsoft's world — Microsoft 365 Copilot, backed by the Office ecosystem, had virtually no competition.

Now OpenAI is entering the ring directly.

ChatGPT Work vs. Microsoft 365 Copilot — this will be the most exciting showdown in enterprise AI over the next two years. On one side, the strongest model capabilities and cutting-edge agent technology. On the other, the most mature office ecosystem and the broadest enterprise customer base. Each side has its advantages; each side has its weaknesses.

OpenAI's advantages: leading model and agent capabilities, fast product iteration, no legacy baggage. Disadvantages: insufficient enterprise services experience, no office software foundation, safety and compliance capabilities need catching up.

Microsoft's advantages: the unbreakable Office ecosystem, deep enterprise customer relationships, mature safety and compliance systems. Disadvantages: slow product iteration, agent capabilities less flexible than OpenAI's, high internal coordination costs.

The outcome of this showdown is hard to call right now. But one thing is certain: the enterprise AI agent market will be the biggest growth market in the AI industry going forward. Whoever wins this market becomes the next king of the AI era.

From conversation to work, from tool to colleague, from consumer to enterprise — every step OpenAI takes lands on an inflection point of industry evolution. ChatGPT Work is just the beginning. The real show is still ahead.

See you tomorrow.