InertiaRSS Track and read blogs, news, and tech you care about
Read Original Open in InertiaRSS

Recommended Feeds

博客园 - 司徒正美
V
V2EX
T
Tailwind CSS Blog
有赞技术团队
有赞技术团队
aimingoo的专栏
aimingoo的专栏
Apple Machine Learning Research
Apple Machine Learning Research
IT之家
IT之家
Blog — PlanetScale
Blog — PlanetScale
A
About on SuperTechFans
月光博客
月光博客
T
The Blog of Author Tim Ferriss
宝玉的分享
宝玉的分享
Martin Fowler
Martin Fowler
博客园 - 聂微东
The GitHub Blog
The GitHub Blog
V
Visual Studio Blog
WordPress大学
WordPress大学
酷 壳 – CoolShell
酷 壳 – CoolShell
Engineering at Meta
Engineering at Meta
GbyAI
GbyAI

阮一峰的网络日志

科技爱好者周刊(第 396 期):互联网通信的替代方案 科技爱好者周刊(第 396 期):互联网通信的替代方案 - 阮一峰的网络日志 科技爱好者周刊(第 395 期):软件开发的第三种方式 科技爱好者周刊(第 395 期):软件开发的第三种方式 - 阮一峰的网络日志 科技爱好者周刊(第 393 期):脑腐状态 科技爱好者周刊(第 392 期):axios 投毒与好莱坞式骗术 科技爱好者周刊(第 391 期):AI 的贫富分化 科技爱好者周刊(第 390 期):没有语料,大模型就是智障 套壳中国大模型撑起500亿美元估值?扒一扒 Cursor 的"套壳"疑云 科技爱好者周刊(第 389 期):未来如何招聘程序员 科技爱好者周刊(第 388 期):测试是新的护城河 零安装的"云养虾":ArkClaw 使用指南 科技爱好者周刊(第 387 期):你是领先的 科技爱好者周刊(第 386 期):当外卖员接入 AI 字节全家桶 Seed 2.0 + TRAE 玩转 Skill 科技爱好者周刊(第 385 期):马斯克害怕中国车企吗? 智谱旗舰 GLM-5 实测:对比 Opus 4.6 和 GPT-5.3-Codex 科技爱好者周刊(第 384 期):为什么软件股下跌 科技爱好者周刊(第 383 期):你是第几级 AI 编程 Kimi 的一体化,Manus 的分层 科技爱好者周刊(第 382 期):独立软件的黄昏 AI native Workspace 也许是智能体的下一阶段 科技爱好者周刊(第 381 期):中国 AI 大模型领导者在想什么 科技爱好者周刊(第 380 期):为什么人们拥抱"不对称收益" 科技爱好者周刊(第 379 期):《硅谷钢铁侠》摘录 我如何用 AI 处理历史遗留代码:MiniMax M2.1 升级体验 科技爱好者周刊(第 378 期):预测是新的互联网热点 科技爱好者周刊(第 377 期):14万美元的贫困线 科技爱好者周刊(第 376 期):太空数据中心的争议 科技爱好者周刊(第 375 期):一扇门的 Bug 终于有人做了 Subagent,TRAE 国内版 SOLO 模式来了 科技爱好者周刊(第 374 期):6GHz 的问题 VS Code 使用国产大模型 MiniMax M2 教程 科技爱好者周刊(第 373 期):数据模型是新产品的核心 国产大模型接入 Claude Code 教程:以 Doubao-Seed-Code 为例 科技爱好者周刊(第 372 期):软件界面如何设计 大模型比拼:MiniMax M2 vs GLM 4.6 vs Claude Sonnet 4.5 科技爱好者周刊(第 371 期):一个乐观主义者的专访 科技爱好者周刊(第 370 期):正确的代码高亮 错误处理:异常好于状态码 科技爱好者周刊(第 369 期):Tim 与罗永浩的对谈 科技爱好者周刊(第 368 期):不要这样管理软件团队 一天之内,智谱和 Anthropic 都发了最强编程模型 科技爱好者周刊(第 367 期):Nano Banana 的几个妙用 科技爱好者周刊(第 366 期):旧金山疯狂的 AI 广告 科技爱好者周刊(第 365 期):流量变现正在崩塌 科技爱好者周刊(第 364 期):最难还原的魔方 科技爱好者周刊(第 363 期):最好懂的神经网络解释 科技爱好者周刊(第 362 期):GitHub 工程师谈系统设计 科技爱好者周刊(第 361 期):暗网 Tor 安全吗?
An Initial Exploration of AI for Office Use: Koushi Space
阮一峰 · 2025-04-21 · via 阮一峰的网络日志

I. The Trend of AI

Let's ask everyone, what is the trend of AI products?

I mean, what kind of products have the greatest opportunity.

Currently, there are countless AI products, most of which are fleeting and don't last long.

In my opinion, the most promising products are neither Q&A AI for the general public or content generation AI, nor programming AI for developers (API users).

The real trend is office-related AI.

The office market for businesses and individuals is enormous and real.

Whoever can integrate AI into documents, meetings, decision-making......, will get the real benefit. Just like Office software is the king of business software, enterprise-level office AI will also be the king of AI sales.

II. The Current State of Office-related AI

However, there are currently very few AI tools for office use. It's not just about winning the market; even products with names are scarce.

The reason lies in part in the current capabilities of AI, which still don't meet the safety, stability, accuracy, and efficiency required for enterprise-level software.

Secondly, the exact product form of office AI is still being explored.

I have been very interested in this field and have a strong interest in new office AI tools.

Next, I'll introduce a new product, "Button Space", which I've been testing out recently. Let's see how it uses AI to complete office tasks.

III. Button Space

First, let me say this: Button (coze.cn) is a domestic platform that provides a low-code environment based on a browser to build AI applications, with a free usage quota.

I have been using it for a long time and used to write articles introducing its workflow mode : on a graphical interface, use a mouse to arrange AI workflows to generate standalone applications.

The workflow mode claims to require no coding, making it accessible to beginners, but users are recommended to have some programming knowledge, so there is still a certain threshold.

Therefore, now Coze has launched "Coze Space," truly barrier-free and accessible to complete tasks.

Its biggest feature is that it internally automatically calls various Agents (intelligences), without user intervention, to complete various tasks . You can imagine it as an "Agent auto-caller."

For users, with the underlying Agent capabilities, it can not only answer questions but also solve problems (tasks), thereby becoming your office assistant and work partner.

As its slogan goes: "Start your work with Agent."

Note that it now requires an invitation code, which can be obtained from the Kouzi WeChat official account and the Kouzi Space official website (space.coze.cn).

Section Four: Interface

The URL for Kouzi Space is space.coze.cn, and you can use it by accessing it.

The interface is simple: the left side is the task list, and the right side is a dialog box for inputting new tasks.

When executing tasks, the default mode is "Exploration Mode," where the AI automatically completes each step, making it faster.

You can also change it to "Planning Mode" to display the intermediate steps of AI thinking, making it easy to adjust at any time.

Besides that, there's nothing else to set up, which is very intuitive.

V. Task Examples

Below are some of my usage examples, all office-related tasks.

Everyone take a look, how well it performs, like a skilled intern proficient in various skills.

5.1 Writing a Research Report

The most common office task is definitely writing documents. I had the AI write the following research report.

I need a research report on the development of the Shanghai tea drink industry in recent years, as well as an analysis of investment opportunities, including industry development, popular products, and more.

Note that for document-related tasks, it's best to specify the output format, otherwise the generated content will be displayed in Markdown format on the conversation page, which is not convenient to use.

I usually add this sentence at the end of the prompt.

Create a visual web page at the same time.

Once it starts running, it will break down the tasks and complete them step by step.

Finally, it gives the generated markdown files and web files.

You can open the webpage in a new window for preview. Below is the webpage it generated, which is rich in both images and text, containing data as well as arguments.

The entire report is divided into four parts, which can be increased or decreased as needed, for personal use or submission, overall there are no problems.

5.2 Multiple output formats

As mentioned before, the Button Space comes built with various Agents.

Besides the Agent for generating web pages, there are also Agents for generating Office documents, PDFs, and Lianxiu documents, all of which can be used. Below is an example.

Help me compare T3, Didi Chuxing, and Gaode, as a novice ride-hailing driver in Chongqing, which platform is most recommended, and explain the detailed reasons. Output a Lark document.

Since Lark documents have a layout, when actually generating them, it's also from a text report to a web page, and then the system prompts you to copy the web page content into Lark.

If outputting a PDF file or a PowerPoint PPT file, the system will directly provide a file download.

I am planning a trip to the southern Anhui region, including Huangshan and other local scenic spots. Please conduct a detailed study of the itinerary, transportation routes, scenic spot introductions, tickets, accommodation, and local cuisine to form a complete itinerary plan in PPT format.

It generated a 28-page PPT file, which can be directly used for presentations after downloading.

5.3 Table Capabilities

For office AI, it must have table processing capabilities.

Generate a table containing the constituent stocks of the SSE 50 Index and their latest closing prices.

By default, the table will be displayed on the web page after generation.

The system will also provide a CSV file for download.

If you want comprehensive spreadsheet functionality, you can import the CSV file into spreadsheet software.

5.4 Other Features

The Agent space of KEG includes many agents, and it can also call Gaode Maps, generate web games, and more.

Generate an interactive learning website to help me learn the OKLCH color function in CSS.

The above is an example of a web mini-game, and the following is an example of calling a map.

Analyze the locations of all Luckin Coffee stores in the Shanghai Bund area using Gaode Maps and create a visualizable web page for me.

If the built-in Agent is not sufficient to meet your needs, you can also set it up to connect to various MCPs to expand its capabilities.

5.5 Expert Systems

Finally, it also comes with an expert system, which is an advanced professional knowledge base, providing deep analytical capabilities.

Currently, two expert systems are built-in: User Research Expert and A-share Observer Assistant.

The former is used for user surveys, interviews, and generating analytical reports; the latter is used for analyzing self-selected stocks and the overall market, providing research and analytical reports.

Deep analysis and processing tasks require specific professional knowledge, so you can try them.

VI. Summary

After initial use, I have a very positive evaluation of Koushi Space, an extremely useful office assistant and AI intern, , which indeed improves office efficiency.

I think, its design philosophy is very sound. It automatically calls various Agents, which can both expand capabilities and eliminate the learning curve, while also possessing deep research capabilities.

What's more难得 is that its completion is quite good, without encountering any obvious bugs, making it suitable for actual office use.

In summary, as an AI product for office use, its "work companion" model, I believe, is both usable and feasible.

Going forward, I expect a large number of office AI products to emerge. Given its importance, I will introduce and review more of them.

(完)