I wrote a book.
Not about how to use AI.
Not about prompts, models, or the latest agent framework.
It is about a harder question:
When AI starts doing real work inside an organization, what has to change around it?
A thread on why I wrote The Accountable Firm. 🧵
Source: OpenAI’s Nov 10 announcement. Businesses in ChatGPT is in beta; initial Business Agents access is limited to eligible US ecommerce and online-service businesses. Business analytics are aggregated, not access to people’s conversations.
https://t.co/IKFH87nBQn
Uncle J · AI & Organizations
Your next marketing task may be fixing the product facts.
With Businesses in ChatGPT, verified companies can update information such as prices and availability. That makes an old pricing page an operational problem, not just untidy copy.
For a one-person business, I’d first check four things: current price, who the product is for, what it cannot do, and refund terms. Give each a source and an owner.
Claiming a profile is not a ranking guarantee. Clear facts are still worth maintaining.
来源:Anthropic 11 月 10 日介绍的 Claude Data。认证查询和 live sharing 仍为 beta;认证功能本身不会替团队决定指标口径。以上是按文档提出的管理建议,未实测。
https://t.co/D7Nnb2xegh
Uncle J · AI & Organizations
两个部门都让 AI 算“活跃客户”,答案不一样,先别急着换模型。
一个算登录过,一个算付过钱。查询都可能没错,会上却没法比较。
Claude Data 新增的认证查询,让数据团队先审核查询,再让业务复用。我的建议是先挑三个会上常争的指标,把定义、查询负责人和修改条件写清楚。
图表生成得越快,越得让人看得见它到底在算什么。
Lower costs mean more room to iterate. The FileFlow demo went from 7 to 12 seconds just to give the preview and confirmation room to breathe. Is the negative markup a launch-only offer, and does it apply across resolutions for both models?