AI Agents: The Integrated Evolution from 'Knowledge' to 'Data'

In more and more enterprises and organizations, AI is no longer just a chatbot.

People want it to help them actually get work done—check policies, read materials, analyze finances, generate reports.

But in reality, we often find:

Documents are documents, databases are databases,
AI can answer text questions but can't read tables;
can search files but can't query business data.

This is precisely the biggest gap that enterprise "AI Integrated Platform" aims to solve.

---

The Key to Integrated Agent: Knowledge Base × Database Fusion

From a technical perspective, AI's two core application scenarios in enterprises are actually fragmented:

Large models themselves are very powerful, but enterprise knowledge and data often exist in layers. Only when these two parts are combined can AI truly understand the enterprise's "full picture."

Therefore, AskTable and other AI platforms (like BetterYeah, Dify, RAGFlow, etc.) form a complementary ecosystem:

• •Knowledge base platforms: Handle unstructured information like policies, systems, and technical documents;
• •AskTable: Handle structured data like databases, reports, and business metrics.

Together, they constitute a unified agent hub within the enterprise.

---

Real Cases: Two Industries, Two Implementation Models

1. Education Industry: AI Assistant Makes Data Services More Accessible

Inside Shaanxi Normal University, AI is deployed as a "Smart Campus Assistant," where teachers and students can directly converse within the system, for example:

• •"Help me check the approval status of research projects."
• •"What are the details of my campus card consumption last month?"

Behind the scenes, this is actually completed by the collaboration of two types of agents:

• •Knowledge-type queries: Handled by the knowledge base AI platform providing document-based models;
• •Data-type queries: AskTable parses semantics, generates SQL, and queries business databases in real-time.

AskTable automatically returns results as tables or charts,
such as research projects by category distribution, monthly consumption trends, etc.
Operations that previously required manually exporting Excel and then analyzing can now be completed in seconds.

Teachers can immediately view research project progress,
students can also query consumption bills and research information,
everyone can directly "ask for data" without having to "find someone to get data."

---

2. Engineering and Infrastructure Industry: From Data Management to Semantic Understanding

Inside a central state-owned enterprise (large infrastructure enterprise), AI is integrated into the digital management system.

The platform contains multiple functional modules:

• •General Q&A: Common sense questions for employees, answered directly by large models;
• •Enterprise knowledge base: Driven by knowledge base platform, supporting file retrieval and policy Q&A;
• •Business queries: Handled by AskTable for the most critical business data access.

Project managers only need to ask in one sentence:

• •"Who is the design unit for a certain building"
• •"In the Y underground parking garage, who is the manufacturer of the screw-type low-energy heat pump unit"

AskTable will automatically parse semantics, generate query statements, and fuse with knowledge base results, achieving natural connection from "asking knowledge" to "asking data."

In the end, business parties can obtain both document information and database results in a unified interface.

AskTable becomes the structured data brain under the knowledge base system.

---

Technical Architecture: Unified Entry, Multi-Agent Collaboration

The core idea of this hybrid architecture is:

"The user talks to only one agent, while multiple agents collaborate behind the scenes to complete the task."

The system logic is as follows:

• •User asks question → Main Agent judges question type;
  • •If knowledge-type → Calls RAG Engine;
  • •If data-type → Calls AskTable to query database;
• •Finally → Unified integration and output of answers, tables, and charts.

AI's identity recognition and permission control for users runs through the entire process, maintaining flexible scaling while ensuring enterprise-level security and control.

---

Why This Matters

Because an enterprise's "knowledge" and "data" have never been in one system.

An AI that only understands text is not enough to support decision-making;
an AI that only queries tables cannot answer questions about policies and logic.

A true agent system must integrate these two.

---

AskTable: The Professional Engine for Data Agents

AskTable focuses on structured data semantic understanding and access:

• •Automatically identify business semantics, generate SQL queries;
• •Support multiple data sources (DaMeng, TDSQL, MySQL, PostgreSQL, SQLServer, StarRocks, etc., more than 20 types);
• •Built-in permission control, field mapping, entity fuzzy matching mechanism;
• •Seamlessly integrate with any knowledge base or agent platform.

It can either serve as an independent "data analysis assistant",
or serve as a "structured intelligent engine" for knowledge base platforms,
working with platforms like BetterYeah, Dify, RAGFlow to build a complete enterprise agent ecosystem.

---

Conclusion: AI's Next Step is Not Being Smarter, but More Complete

Making AI capable of not only querying documents but also querying databases, analyzing business, generating decisions,
this is the true implementation path for enterprise intelligence.

The combination of AskTable and knowledge base platforms
is precisely the key step in moving AI from "asking knowledge" to "asking data."