The access to DeepSeek for government affairs systems in many places has brought these changes to the "AI civil servant" job

The 12345 hotline in many places across the country announced the access to DeepSeek, and the Yiwu boss used DeepSeek to sell small commodities. Grasp information in seconds, and novices can also generate videos of foreign languages. How can artificial intelligence promote service upgrades and bring more business opportunities?

Does of AI and digital intelligence employees in Futian District, Shenzhen, Guangdong have been "on-job". It can be said that artificial intelligence is continuing to promote the quality and upgrading of various services such as government services.

At present, the 12345 hotlines in Liaoning, Zhanjiang, Nanchang and other places have announced that they have been connected to DeepSeek. What new experience can the integration of AI technology and 12345 hotline innovation bring? We can learn about one or two through a specific example.

For example, if you want to know where to apply for a reissue ID card without a municipal household registration, and what materials should you submit? So I called 12345 to check. Before you make this call, the relevant policies have been imported with one click, and artificial intelligence automatically resolves the terms and builds an associated knowledge graph. When you ask a question, AI can analyze semantics in real time, automatically match policy basis and solutions in the knowledge base, and answer your questions. When you hang up the phone, the artificial intelligence will automatically generate a structured summary. The operator can export and synchronize it to the return visit system with one click for subsequent statistical return visits and other uses.

What new changes will be brought about by combining 12345 hotlines in various places and big models? How will artificial intelligence continue to promote the upgrading of government services?

On February 20, Jiangmen, Guangdong announced that it has launched the DeepSeek-R1 general capability in the domestic environment based on the government cloud, and has been applied in the 12345 government service convenience hotline, intelligent government official documents layout, intelligent government service guidance and other scenarios. Liang Yu, deputy director of Jiangmen Municipal Government Services and Data Management Bureau, introduced that in terms of government service hotline, the Deepseek big model was tried in many aspects such as intelligent question-and-answer, intelligent form filling, and intelligent policy interpretation. In addition to the telephone side still being answered manually, the question-and-answer on the Internet side has been handed over to the trained big model, and then supplemented by manual: Liang Yu, deputy director of Jiangmen Municipal Government Services and Data Management Bureau: The 12345 we usually say is a telephone end, but in fact we also have a network end. At present, the first one we went online is on the network end. When citizens asked questions and answers on our network end, because this was a text-like appeal. Basically, 70% of the work orders in consulting categories are accounted for. In the traditional model, intelligent search is used to assist in humans. We search for questions from citizens in our knowledge base. After searching for answers, we will provide feedback through manual methods. The efficiency is not so high. After introducing the big model, it will have a good understanding of the questions from citizens and based on its understanding, we can search for matching answers in the knowledge base we provide.

Of course, the big model of localization is not like we use web pages or APPs, but has been trained locally in Jiangmen, making it a "Jiangmentong".

We imported 104,500 pieces of data from the hotline knowledge base into the big model for training. We imported our knowledge base directly into the big model, which is equivalent to feeding it the corpus and asking it to understand the answers to the knowledge base. The right human beings will review and supplement it.

Liang Yu told reporters that they are trying to use big models to analyze backend data, hoping to discover the trend of problems reflected by citizens: