Next-generation BI tool Codatum launches beta version of AI data analysis agent 'Codatum Agent'
CODATUM Inc. has launched the beta version of 'Codatum Agent', an interactive data analysis agent that autonomously handles everything from SQL generation to visualization using natural language.
📋 Article Processing Timeline
- 📰 Published: April 15, 2026 at 18:00
- 🔍 Collected: April 15, 2026 at 09:31
- 🤖 AI Analyzed: April 19, 2026 at 14:03 (100h 31m after Collected)
CODATUM Inc. (Headquarters: Tokyo, CEO: Naoki Shibayama) has announced the launch of the beta version of a new feature, 'Codatum Agent', within its next-generation BI tool 'Codatum'.
Codatum Agent (Interactive Data Analysis)
'Codatum Agent' is an interactive data analysis agent that autonomously completes everything from table searches to SQL construction/execution and chart generation, simply by requesting analysis in natural language. By continuously learning team knowledge, its accuracy steadily improves. Its code-first design, which allows professionals to review and modify, ensures reliable analysis.
Feature Overview
Interactive Analysis - Completes exploration to visualization entirely with natural language
Simply request in natural language, and AI autonomously handles everything from table exploration to SQL construction, execution, and chart generation. It automatically fixes errors during the process, and analysis is completed with instructions like 'aggregate monthly sales and make a graph' without writing SQL. By specifying Notebooks or catalogs as context, analysis based on specific preconditions is also possible.
Notebook Integration - Reports are built while conversing
SQL and charts generated through conversation with the Agent can be directly saved and shared as Notebooks (document format / dashboard format). Appending to existing Notebooks is also done simply via conversational instructions. Analysis results do not become siloed but accumulate as team knowledge.
Learning Capability - Gets smarter the more you use it
It learns from the insights of created Notebooks, catalog metadata, and agent contexts set by professionals (rules specific to workspaces/teams/individuals/projects), continuously improving the Agent's accuracy.
Code-First - Reliable because professionals can review
Because both SQL and charts remain as code, the Agent's entire thought process is visualized. Professionals can review and modify them, and the detected and corrected hallucinations are reflected in its learning. AI models can also be switched depending on the scale and accuracy requirements of the analysis (paid option).
Use Cases
'Graph the sales trends by channel from last week's order data' -> Automatically identifies tables, generates/executes SQL, and adds charts.
'Add a summary of monthly KPIs to this Notebook' -> Appends while referencing existing Notebooks.
'Investigate the cause of the drop in conversion rate' -> Investigates, aggregates, and reports results across multiple tables.
'Create a new sales report Notebook' -> Automatically creates the Notebook from structure to completion.
Comment from Masatake Yamoto, CEO of 10X Inc.
At 10X, multi-dimensional analysis is required daily in our AI ordering business. With traditional BI tools, the workflow from analysis to sharing and discussion was fragmented, causing data utilization to become isolated to certain members. After implementing Codatum, simply by verbalizing a hypothesis, AI completes everything from SQL generation to chart creation and documentation end-to-end, making it easy to tackle analysis we would have previously given up on. Especially with Codatum Agent, the speed and quality of analysis have significantly improved. I strongly feel that the democratization of data utilization is progressing steadily, making it an indispensable tool for teams aiming for data-driven decision-making.
*Reference Info: Yamoto discusses the use of Codatum in detail on the Podcast 'Zero Topic' #340.
Availability and Conditions
- Launch Date: April 15, 2026
- Eligible Plans: Business Plan and above (Optional feature)
- Service URL: https://codatum.jp/agent
- Support Document URL: https://docs.codatum.jp/codatum-ai/ai-agent
About the next-generation BI tool 'Codatum'
Codatum is a next-generation BI tool that utilizes AI and Notebooks to enable all team members to handle data freely. By combining AI, SQL, and no-code, anyone can...
Codatum Agent (Interactive Data Analysis)
'Codatum Agent' is an interactive data analysis agent that autonomously completes everything from table searches to SQL construction/execution and chart generation, simply by requesting analysis in natural language. By continuously learning team knowledge, its accuracy steadily improves. Its code-first design, which allows professionals to review and modify, ensures reliable analysis.
Feature Overview
Interactive Analysis - Completes exploration to visualization entirely with natural language
Simply request in natural language, and AI autonomously handles everything from table exploration to SQL construction, execution, and chart generation. It automatically fixes errors during the process, and analysis is completed with instructions like 'aggregate monthly sales and make a graph' without writing SQL. By specifying Notebooks or catalogs as context, analysis based on specific preconditions is also possible.
Notebook Integration - Reports are built while conversing
SQL and charts generated through conversation with the Agent can be directly saved and shared as Notebooks (document format / dashboard format). Appending to existing Notebooks is also done simply via conversational instructions. Analysis results do not become siloed but accumulate as team knowledge.
Learning Capability - Gets smarter the more you use it
It learns from the insights of created Notebooks, catalog metadata, and agent contexts set by professionals (rules specific to workspaces/teams/individuals/projects), continuously improving the Agent's accuracy.
Code-First - Reliable because professionals can review
Because both SQL and charts remain as code, the Agent's entire thought process is visualized. Professionals can review and modify them, and the detected and corrected hallucinations are reflected in its learning. AI models can also be switched depending on the scale and accuracy requirements of the analysis (paid option).
Use Cases
'Graph the sales trends by channel from last week's order data' -> Automatically identifies tables, generates/executes SQL, and adds charts.
'Add a summary of monthly KPIs to this Notebook' -> Appends while referencing existing Notebooks.
'Investigate the cause of the drop in conversion rate' -> Investigates, aggregates, and reports results across multiple tables.
'Create a new sales report Notebook' -> Automatically creates the Notebook from structure to completion.
Comment from Masatake Yamoto, CEO of 10X Inc.
At 10X, multi-dimensional analysis is required daily in our AI ordering business. With traditional BI tools, the workflow from analysis to sharing and discussion was fragmented, causing data utilization to become isolated to certain members. After implementing Codatum, simply by verbalizing a hypothesis, AI completes everything from SQL generation to chart creation and documentation end-to-end, making it easy to tackle analysis we would have previously given up on. Especially with Codatum Agent, the speed and quality of analysis have significantly improved. I strongly feel that the democratization of data utilization is progressing steadily, making it an indispensable tool for teams aiming for data-driven decision-making.
*Reference Info: Yamoto discusses the use of Codatum in detail on the Podcast 'Zero Topic' #340.
Availability and Conditions
- Launch Date: April 15, 2026
- Eligible Plans: Business Plan and above (Optional feature)
- Service URL: https://codatum.jp/agent
- Support Document URL: https://docs.codatum.jp/codatum-ai/ai-agent
About the next-generation BI tool 'Codatum'
Codatum is a next-generation BI tool that utilizes AI and Notebooks to enable all team members to handle data freely. By combining AI, SQL, and no-code, anyone can...