We interviewed 10 workplace overachievers to see which AIs they are actually using.

36Kr "Workplace Bonus" (ID: WorkplaceBonus)

With DeepSeek setting off a new wave of AI craze in 2025, some enterprises have started the "Efficiency Revolution". And perhaps the only way for workers to save themselves is to keep up with the times and first carry out a round of "Self-Enhancement".

On February 7, there were market rumors that Shanghai Shangmei Co., Ltd., the parent company of Hanhoo, intends to use AI to replace manual labor. The CEO issued an "employee reduction ratio" within the department, with the legal department eliminating 50%, the customer service department 95%, the new product innovation center 70%, the content innovation department 80%, leaving only a few AI-savvy employees. Not only that, but employees are also required to train the ability to communicate with AI through the mouse and language, so the keyboard is to be cancelled. After the news spread, CEO Lv Yixiong refuted the rumors in his Moments, stating that it is not a layoff, "Some places add people, some places reduce people, to make the operation more scientific".

The application of AI in work is not a rare thing. However, although various applications of AI are increasingly widespread, the level of AI usage among workplace people is uneven - this has also made the ability to use AI well a new dimension that widens the individual career competitiveness.

6 Types of AI Tools Indispensable for Workplace People ╱ 01

It Turns Out AI Can Be Used This Way ╱ 02

Written in the End ╱ 03

 

Wang Han is an experienced algorithm engineer and one of the first batch of users to try using ChatGPT to write code. At the beginning, the effect was not ideal, and it was not as good as writing it himself. But now AI has become an indispensable part of his work. The turning point started when he met Claude: He found that by breaking down the project tasks into individual modules that can run independently and communicating the requirements to Claude with a programmatic thinking, the code obtained can have an 80% to 90% completion rate.

"Writing prompt words" is the part that can most widen the gap in the level of AI usage in practical operations, and it has also become a new skill that workplace people need to spend time honing. In Wang Han's eyes, generative AI is like a newcomer to the industry, but one who can understand with a little guidance. As long as you are patient in guiding, it can bring surprises. "I often help my colleagues optimize prompt words. Many people give AI instructions that are still too broad."

After reviewing the recent data, we have discovered the following several important pieces of information [1]:

1. More than 60% of workers use AI technology to improve work processes, automate daily tasks, and perform data analysis, etc.

2. The usage proportion of various AI products among workplace people: Text writing (85%), Image generation (53%), Video processing (54%), and PPT production (39%), etc.

3. From January to October 2024, The number of positions that explicitly require the use of AI tools has increased by 68% year-on-year, especially in fields such as product, technology, design, and content production.

4. Professions with a high AI impact index, such as editing, translation, customer service, software and hardware development, etc., Although the overall recruitment demand shows a downward trend, the requirements for candidates' AI skills are increasing.

Workplace people who are proficient in using AI skills can stand out more in the competition. But the reality is that many people still only stay at the basic functional usage and fail to give full play to the role of AI - this gap stems from information asymmetry, and it will further lead to an intensification of information asymmetry. A relevant researcher from an AIGC research institute told "Workplace Bonus" that they found that most users simply tried it once and concluded that "AI is not good". "Ordinary users often do not give AI much patience, but this also determines that different people use AI, and the effects are also very different."

In order to break this information barrier and bring more inspiration to workplace people. "Workplace Bonus" conducted in-depth interviews with 10 experts who use AI in their daily work, and we summarized 6 types of practical AI tools for workers. There are not only further comparisons and ability explorations of the familiar tools of the "old players", but also some relatively unpopular but useful tool recommendations.

The review logic of this article mainly focuses on the basic usage scenarios. However, when it comes to the application of AI in the workplace, in fact, more "integrated and combined" gameplay and applications can be generated by combining the vertical scene requirements of different industries according to different job functions (such as product, operation, marketing, strategy consulting, human resources, finance, and law, etc.). Here is an event preview: On February 22, we will hold a Spring Talent Salon "Using AI to Benefit Our Careers" in Beijing, and the " On-the-Job Chapter " session will bring offline sharing of related content.

● Note: Due to venue and topic limitations, only less than 100 people are allowed to participate

Without further ado, let's take a look at the inspiration and skills of "Workplace Kings" in taming AI.

[1] Information source: 51job survey data, Liepin's "2024 Extraordinary Insights Report", and the "Research on the Potential Impact of AI Large Models on China's Labor Market in 2024" jointly released by the National Development Institute of Peking University and Zhaopin.

● Let's start with a summary chart, and we'll explain in detail later

6 Types of AI Tools Indispensable for Workplace People

As an application with over 100 million users within 7 days of its launch and over 20 million daily active users within 20 days, DeepSeek is often in a "server busy" state recently.

36Kr "Workplace Bonus" (ID: WorkplaceBonus)

Fortunately, many third-party APPs have taken prompt actions to achieve local deployment of DeepSeek. For example, 360 Nano Search, upon opening the APP, you can see the DeepSeek conversation entry at the top. After testing, although the answers are not exactly the same as the original, it adds visual and voice content.

● 360 Nano Search

The AI product Flowith (https://flowith.io/blank), which focuses on the thinking canvas function, has also integrated DeepSeek. Flowith can not only produce smoothly during DeepSeek's "frequent off-duty periods", but also can call the large models of other companies for Q&A.

● Flowith Official Website

These applications or webpages do not integrate the DeepSeek product itself, but only locally deploy open-source models, so the application effect may not be as good as the product itself. And in order to optimize the product experience, DeepSeek has built-in many prompt words, so users do not need very professional prompt word content to obtain "DeepSeek-flavored" content.

Now, many companies have locally deployed the DeepSeek large model, such as Geely Automobile, 360, Silicon Mobility, and Huawei. To a certain extent, DeepSeek has also promoted the integration of generative AI into other fields.

36Kr "Workplace Bonus" (ID: WorkplaceBonus)

In the daily work of workers, Baidu Wenku is one of the earliest to launch AI functions and charge for them, and Quark APP has also updated functions such as AI PPT, AI search, AI writer, and AI search. Feishu has Feishu Miao Ji; WPS also has an AI assistant, Tencent Meeting has launched an AI assistant for meetings, and so on.

● Quark APP

Currently, the AI assistants for organizing meeting minutes, such as Feishu Miao Ji (mainly for voice-to-text conversion) and the AI assistant entry of Tencent Meeting, are easy to find. Feishu Miao Ji has a free quota per month (300 minutes), and if the quota is exceeded, a fee is required; to enable the small assistant in Tencent Meeting, a direct payment is required (¥88/month).

● Feishu Miao Ji

● Tencent Meeting

But just as Baidu's "Wen Xiaoyan" assistant started charging as soon as it was launched, after paying, users expect to get more efficient generation results. However, the uncertainty of the output of generative AI, coupled with the double pressure of users' not-so-mature usage mentality, often fails to meet the satisfaction range of users, and the paid content of major companies is often criticized. This also brings opportunities to other applications in the market.

● Daily Office Work

In terms of processing text and data-related work, ChatGPT, Claude, Doubao, Kimi, Tongyi Qianwen, and DeepSeek, these top players are in a fierce competition.

ChatGPT and Claude cannot be directly used in China. The former has excellent comprehensive capabilities, while the latter's copywriting and programming capabilities are slightly better than ChatGPT. Relevant people shared with "Workplace Bonus" that some creators directly trained the AI in Claude to write public account articles (the process will be shared in the next part). The trained Claude assistant can output an article in a few minutes, not only can it be "taken for granted", and one article has received a reading volume of 100,000+.

But the long-standing problem that bothers everyone about Claude is that it "loves to ban accounts". This problem can be solved through Poe (a software that integrates AI assistants). Poe's free quota is approximately 3-5 generations per day, and the upgrade membership fee is around more than one hundred yuan per month.

36Kr "Workplace Bonus" (ID: WorkplaceBonus)

The most popular domestic general assistants are Doubao under ByteDance, Kimi under the big model unicorn Dark Side of the Moon, and Tongyi Qianwen of Alibaba, which is deeply engaged in technology. According to the overall feedback from netizens and users, Doubao is better at writing with a humanistic touch than Kimi; Doubao and Tongyi Qianwen are better at official document writing; Kimi is more rigorous in literature retrieval, and the cited literature is real, while Doubao is more prone to hallucination phenomena.

By the way, for literature retrieval, some people also recommend using Bohrium (https://bohrium-doc.dp.tech/docs/).