Google's ToC assistant makes a splash, and Feishu's ToB "Knowledge Q&A" follows suit | Exclusive from Intelligence Emergence

Author | Deng Yongyi

Editor | Su Jianxun

Early in the morning on May 21st, Beijing time, Google once again made a splash at its annual I/O Conference - the AI mode of Google Search was officially launched. Among them, the most eye - catching feature is Personal Context.

Simply put, this is a memory function. After obtaining user authorization, Google's large - model assistant Gemini can access your search history, email, bookmarks, cloud storage, and other data, thus providing more user - tailored Q&A services. For example, it can retrieve the time and location of a specific schedule from an email and give more personalized suggestions based on the weather and the user's location preferences.

However, you don't have to wait for Google. An enterprise - version Context will soon be available in China.

According to exclusive information obtained by "Intelligent Emergence", Feishu is about to launch a new AI product - Feishu Knowledge Q&A.

"Feishu Knowledge Q&A" is an enterprise - specific AI Q&A tool based on enterprise knowledge, which can be accessed from the left toolbar of Feishu. When users encounter problems at work, they can ask questions to Feishu Knowledge Q&A at any time.

Simply put, in terms of user experience, it is more like an intern who "can remember things". When users encounter problems at work, they can ask it questions at any time. After users set the "knowledge scope" it can access, Feishu Knowledge Q&A can access all information such as messages, documents, knowledge bases, and files, and directly provide accurate answers.

For this reason, AI products similar to enterprise knowledge bases have become a direction that both large companies and startups are vying to enter. Information such as users' data documents and conversations in the workplace are all used as "reference materials" for the model to call.

Taking Feishu Knowledge Q&A as an example, a practical application case is as follows: All the knowledge that users have access to in Feishu will automatically become their knowledge base without manual sorting and uploading. When users ask questions to Feishu Knowledge Q&A, the AI will automatically call these materials.

In this way, there are more practical use cases. According to the product demonstration released by Feishu, there are scenarios such as inquiring about project progress, asking about budget status, finding people, and writing project summaries.

When large models are implemented, the issue of enterprise information confidentiality often arises. In the past, if employees casually fed enterprise information to other general large models, it might lead to information leakage, and the enterprise's unique knowledge and experience might become answers for external users.

How to solve the security and privacy issues? As we understand, the permission management of Feishu Knowledge Q&A is currently based on the permissions of the product's own documents, meetings, group chats, knowledge bases, etc., which are consistent with the user's permissions. Therefore, even if an ordinary employee, a business director, or the CEO asks the same question, the generated answers will be completely different.

"Intelligent Emergence" has learned that currently, this product has completed internal testing with tens of thousands of users, and the web version (ask.feishu.cn) has also been officially launched. Users can search for "Knowledge Q&A" in Feishu or directly visit the website to apply for a trial.

From the web version, Feishu Knowledge Q&A is similar in form to GPT and Doubao with long - term knowledge memory. Users can upload personal data. After uploading, they can use DeepSeek or Doubao to ask questions. The web version also supports online search and knowledge base search and other methods.

△Source: Feishu

After the generational progress of inference models represented by OpenAI o1 and DeepSeek, the accuracy of model output and the hallucination problem can be effectively controlled, which directly affects the implementation direction of AI products. Products such as Tencent ima and Quark have increased their investment in knowledge bases. On the side of startups, products similar to personal knowledge bases such as Youmind and Remio have also had a "small explosion", and the number of users has increased rapidly.

Cover Image Source | AI - generated

Welcome to communicate

Welcome to follow

This article is from the WeChat official account "Intelligent Emergence", author: Deng Yongyi. It is published by 36Kr with authorization.