Smartbi AIChat
The AI model was delivered by Sematic and participated in the list/award selection of the "2024 China Digital Intelligence Industry AI Large Model Pioneer Enterprise" jointly launched by Data Ape and Shanghai Big Data Alliance.
Smartbi AIChat is the latest self-developed agent-based AI Agent, and combines years of accumulated business intelligence (BI) technology and rich industry know-how capabilities with artificial intelligence technology. Through the in-depth training of third-party large models in functions and scenarios, Smartbi AIChat integrates advanced data model calculation, machine learning, and metric management functions to achieve a comprehensive reconstruction of product capabilities. This innovation not only significantly improves the credibility and security of data-driven decision-making, but also accelerates the mining of data value and greatly improves the efficiency of data usage, providing enterprise users with a flexible, easy-to-use, and accurate data analysis experience.
Application Scenarios/User Groups
Application scenarios: Smartbi AIChat can be widely used in the process of digital transformation of various industries, such as enterprise marketing, production, finance, personnel and other business scenarios to meet the needs of data analysis.
User group: Smartbi AIChat products are suitable for enterprise users who face difficulties such as long data analysis links, low timeliness, limited efficiency and flexibility of report production, closed-loop data and difficult applications in the digital operation of enterprises. On the data development side, developers pay more attention to the modeling of data models and the construction of data infrastructure. On the data consumption side, by using Smartbi AIChat, enterprise business personnel can quickly realize data analysis through simple dialogue Q&A, giving each user the ability to "increase analysis". At the same time, for managers, they can control the business status anytime and anywhere, and provide data analysis results that support decision-making quickly, accurately and efficiently. At present, under the guidance of national policies, central state-owned enterprises and financial institutions have relatively rigid AIGC landing needs, and will be the first batch of customer groups.
Product features
Product features:
Smartbi AIChat is based on AI Agent technology and expands capabilities through plug-ins. Complete the answer to an open-ended question by generating Python code; It retrieves privatized data based on Smarti indicator model and data model, sets permissions and data privacy, ensures data security, and supports hundreds of millions of data access. In addition, it provides deep insights into data patterns and trends, adding attribution, predictive analytics, data mining, data interpretation, and more.
Usage:
(1) Prepare data source: support Excel data file import or link to database;
(2) Build Smartbi data model: Smartbi data preparation, setting and establishing table relationships, indicator dimensions, etc.;
(3) Data model adaptability transformation: collect questions according to business needs to guide modeling and testing, ensure the coverage of indicators and guide the naming of dimension indicators as much as possible, and transform the model;
(4) Generate AI graph: save the data model after adaptation transformation, build a model map, set the construction dimension and other configurations;
(5) Use conversational analysis to answer data and generate visualization results in real time;
(6) Continuously optimize the operation and maintenance of questions to improve the overall output accuracy.
Product Advantages:
Smartbi AIChat significantly improves the efficiency and convenience of self-service data analysis through large model technology, and its advantages are mainly reflected in:
(1) Very low learning cost: no need for cumbersome training, users can directly ask questions through natural language, and the system can generate visual results instantly;
(2) Instant business empowerment: Quickly respond to sudden and temporary business problems, quickly solve doubts through direct questions, and accelerate the decision-making process;
(3) Personalized recommendation: Based on the user's questioning habits, the system intelligently recommends and analyzes dimensions and indicators to optimize the user experience;
(4) Multi-terminal support: support multi-platform operation, users can conduct data analysis anytime and anywhere, providing work flexibility;
(5) Data collaboration: It can be embedded in internal communication tools to promote data sharing and team collaboration, and strengthen data consumption capabilities;
(6) Large language model support: Support Alibaba, Deepseek, Baidu and other large models, as well as privatized deployment, ensure data security and privacy, improve the ability to understand semantics and context, and efficiently meet data needs by accurately analyzing user language.
Technical Notes
Smartbi AIChat has realized the transformation of the analysis method with indicator management combined with AI dialogue search as the core, and has made great breakthroughs and innovations in algorithm technology, which is manifested in: framework design according to the large model + Smartbi data model/indicator model, the large model is mainly used for the connection and transformation of natural language and DSL analysis statements, and the Smartbi data model is used as the core infrastructure of data storage and query analysis. The thinking path is as follows:
(Including some anomaly analysis, attribution analysis, predictive analysis)
(1) If the user enters the problem, it is retrieved by Schema Mapper to determine whether the field matches the business knowledge base;
(2) If there is a match, skip the large model parsing step and directly use the indicator calculation formula in the knowledge base to trigger the Smartbi data model for query and analysis;
(3) If there is no match, firstly, RAG retrieval is carried out through natural language questions, combined with the semantic layer of Smartbi data model, and DSL statements are formed by combining with large model semantic parsing.
(4) DSL statements are further filtered and disassembled through the semantic layer to generate simple intermediate definitions of single tables to trigger data processing and query acceleration of the Smartbi data model.
(5) For query scenarios that need to be combined with external information, the large model will determine whether to call a third-party plug-in to assist in completing the query.
Serving customers
A large insurance group, a leading financial brokerage, and a well-known artificial intelligence listed company.
About the business
· Sematic
Founded in 2011, Guangzhou Sematic Software Co., Ltd. (hereinafter referred to as "Sematic Software") focuses on business intelligence and big data analysis software and services, and has developed into a leading business intelligence BI and AI application manufacturer in China. The company has provided one-stop business intelligence platform and BI solutions for more than 5,000 enterprises, serving customers in more than 60 industries such as finance, central state-owned enterprises, manufacturing, etc., and is the manufacturer with the largest number of industry head customers in the field of business intelligence.
Sematic Smartbi
Smartbi is a leading BI brand in China, since its establishment in 2011, it has always adhered to the research and development and innovation of BI products, and is committed to providing one-stop business intelligence solutions for enterprise customers, and has been recognized by 5000+ leading customers in finance, manufacturing, medical, retail, education and other industries.
★ The above project cases submitted and declared by Sematic will eventually compete for the list/award of "2024 China Digital Intelligence Industry AI Large Model Pioneer Enterprise" jointly launched by Data Ape and Shanghai Big Data Alliance.
The list will finally be announced for the first time at the "2024 Enterprise Digital Intelligence Transformation and Upgrading Development Forum - and AI Large Model Trend Forum" held in Beijing on July 24, and an award ceremony will be held