Business Intelligence (BI) aims to help enterprises convert data into value and enhance operational capabilities. Both traditional and self-service BI can only bridge the gap between data and people but cannot eliminate it. The emergence of AIGC has driven the development of intelligent BI, disrupted the way people interact with data, and has the potential to reduce or even bridge this gap.

Furthermore, most conversational BI products currently available in the market focus primarily on data querying and cannot perform in-depth analysis (such as diagnosis, attribution, and prediction). Therefore, we need a product that significantly lowers the usage barrier while enhancing analytical depth tailored to various industries.

On August 8th, Smartbi Software proudly launched an independent AI application, Smartbi AIChat, also known as "Bai Ze".

Bai Ze originates from the classic Chinese mythology "Shan Hai Jing" (Classic of Mountains and Seas), where it is depicted as a mythical creature that knows everything, possesses boundless wisdom and insight, and provides solutions to problems.

Smartbi AIChat boasts exceptional language comprehension and conversion capabilities, understanding various industries and insights from billions of data points. It not only enables free, rapid, accurate, and secure intelligent querying but also delves deep into uncovering hidden causes, values, and trends behind the data, offering practical and effective solutions that demonstrate profound insight.

Let's delve into the functional features and underlying core technologies of Smartbi AIChat.

01. Bai Ze's Wisdom: Demonstrating Intelligent Insight

Smartbi AIChat is incredibly straightforward, allowing anyone to easily gain insights into data through intelligent Q&A, significantly enhancing BI adoption within enterprises. Unlike other intelligent BI products, Smartbi AIChat provides not only descriptive analysis but also diagnostic, predictive, and prescriptive analysis with prompts.

Scenario 1: Natural Language Conversational Querying and Interaction

Smartbi AIChat precisely understands intent and maintains coherence, accuracy, and depth across multiple rounds of conversation. Whether it's managers, data analysts, or business personnel, all can efficiently and accurately obtain business information through natural conversations.

Through relaxed, natural, and casual chats, engage in cordial conversations with your personal digital assistant, who assists you in accessing all desired data and presenting it in any desired format. Even if the digital assistant struggles to comprehend your intent, you can directly correct or guide it to ensure a smooth data analysis process.

Scenario 2: Data Computation Capabilities

When querying data that's not directly accessible in the underlying database, such as year-over-year values, year-on-year growth rates, month-over-month (MoM) changes, MoM growth rates, or cumulative values from the start of the year by quarter or month, Smartbi AIChat excels with its robust time-series intelligent computation capabilities. It supports complex time expressions like "three consecutive years" and "year-to-date accumulations," enabling easy cross-time accurate filtering, aggregation, and comparative analysis through simple queries.

Scenario 3: Attribution, Prediction, and Data Interpretation

Smartbi AIChat supports attribution analysis, automatically identifying and presenting critical influencing factors, enabling users to swiftly pinpoint causes and make decisions.

By simply inputting a query like "Analyze the reasons for the decline in contract value in October 2023," Smartbi AIChat automatically analyzes dimensions like product type, sales channel, customer type, and promotional activities to identify key factors influencing the contract decline.

Leveraging advanced machine learning algorithms and AI technology, Smartbi AIChat facilitates data prediction tasks through simple conversational operations, predicting future data trends. For instance, inputting "Predict the trend of contract value for the next three months." Additionally, the large model can interpret complex data, clearly summarizing data characteristics, flagging anomalies, and even suggesting next steps for analysis.

Scenario 4: Complex Nested Queries

For intricate business scenarios, like identifying departments with consecutive growth in performance within a specified timeframe and top-performing salespeople within those departments, Smartbi AIChat supports time-range queries. For instance, querying "contract amounts from January 2022 to March 2023" and then performing consecutive growth analysis like "find sales departments with contract amounts growing for three consecutive months" based on those results, enabling nested queries for more complex business scenarios.

Currently, Smartbi AIChat's capabilities encompass various data analysis scenarios that business users may encounter daily. This streamlined interaction process brings the vision of "making everyone a data analyst" closer to reality.

02. Bai Ze's Leadership: Stemming from Technological Innovation

Smartbi consistently pursues independent innovation, focusing on big data analysis, machine learning, and AI research, staying at the forefront of technological development.

Most conversational BI products in the market utilize LLM and NL2SQL technologies to convert users' natural language queries into SQL statements for data querying. However, when facing highly complex and reliable data analysis applications, this approach can lead to inaccuracies, insecurity, computational difficulties, and shallow analysis due to large model hallucinations.

To address these issues, Smartbi AIChat integrates cutting-edge AI technologies (RAG + large models + semantic models + AI Agent architecture) with Smartbi's BI capabilities (visual analytics, machine learning, and data modeling), leveraging customers' private domain data to efficiently and precisely identify intent, safely and accurately access data, and conduct in-depth analysis, insights, and future predictions.

First, Professional Semantic Models for Understanding Private Business Data

Smartbi's data models easily integrate multi-source heterogeneous data, enabling complex data computations and enhancing query performance. The indicator model, built upon the data model, constructs a unified self-growing indicator system, preserving industry expertise (Know-How).

Smartbi AIChat leverages data and indicator models as its semantic layer, akin to having a robust, accurate, fast-responding, and secure underlying data engine. This provides the data foundation for Smartbi AIChat's comprehensive industry knowledge.

Second, RAG Retrieval Enhancement for Efficient and Accurate Intent Identification

The introduction of a professional semantic layer prevents large models from making unsubstantiated claims, ensuring safer and more reliable results. However, to further comprehend users' analytical intent, more private business knowledge and user behavior habits must be incorporated. To this end, Smartbi AIChat employs RAG retrieval enhancement technology.

The natural language querying process involves multiple steps:

Firstly, we use Embedding technology to embed indicators and dimensions from the data model into a vector database. Further, we incorporate synonyms, knowledge bases, business rules, and user feedback into the vector database. Then, based on the user's query, we narrow the search scope through vector retrieval, rules, and BERT model preprocessing. Finally, we leverage the general knowledge of large models for precise matching and DSL conversion into Smartbi's model queries to retrieve data.

RAG technology enhances the large model's reasoning abilities and matching efficiency, providing the technical foundation for Smartbi AIChat's comprehensive industry knowledge.

Third, AI Agent for Flexible Expansion of In-Depth Analysis

Most BI tools only address data querying. For deeper anomaly detection, attribution, and predictive analysis, users often rely on specialized machine learning platforms.

Smartbi AIChat employs an increasingly intelligent AI Agent framework. With an AI Agent, it's like having a logical, efficient assistant. Upon receiving a task, it proactively thinks, identifies, plans, breaks down, and executes.

The AI Agent also endows Smartbi AIChat with remarkable flexibility, enabling in-depth analysis by calling upon Smartbi's rich algorithm plugins within its machine learning products.

Fourth, Layered Computational Capabilities for Intelligent Handling of Complex Computations

For complex computations like year-over-year comparisons, "three consecutive months," and attribution, Smartbi AIChat fuses Smartbi's existing computational capabilities through the AI Agent, classifying them after defining computational boundaries. For large-scale statistical computations, it leverages the data model's robust capabilities. For other complex computations, it executes them outside the library using Python.

03. Bai Ze's Excellence: Backed by Industry Experience

With 13 years of experience in the BI industry, Smartbi has amassed over 5,000 customer cases, accumulating valuable knowledge across industries like finance, government, manufacturing, healthcare, and education. This knowledge, carried by indicator models, forms industry Know-How, providing a solid business foundation for Smartbi AIChat's comprehensive industry understanding. Furthermore, supported by semantic models and RAG technology, Smartbi AIChat boasts extensive industry management expertise, offering enterprises comprehensive, efficient, and implementable intelligent BI solutions.

After years of refinement and accumulation, Smartbi AIChat has been successfully implemented in the financial sector and government agencies, and strategic partnerships have been forged with large model vendors. These benchmark customer successes will provide valuable experiences and case studies for more enterprises, driving the healthy development of the entire AI industry.

Currently, Smartbi AIChat primarily targets data consumers (business personnel and managers), enabling deeper analysis through Minimalist and accurate conversational analysis service . As AI technology continues to evolve and find new applications, we aim to expand our offerings to various roles, delivering more advanced features that further help enterprises reduce costs, enhance operational efficiency, and improve decision-making capabilities.