Dataiku Japan Inc. announced on the 14th that it will begin offering new functions that enable the creation and control of AI agents. The functions will be provided on the company’s “The Universal AI Platform.”
In response to this, the company held a press conference, where Dataiku Japan President and Country Manager Yutaka Sato spoke about the challenges of AI implementation, the potential of AI agents, and the company’s efforts in this regard. According to Sato, Dataiku has set a vision of “Everyday AI” since 2018, and “aims to integrate AI into everyday work without treating it as something special.” Under this vision, Sato said, “The key to AI is not just technology, but also organizational change. To that end, we are working to make AI agents a part of everyday life.” The Universal AI Platform supports this effort. Sato said that there are five barriers to AI implementation: “technology,” “tools,” “organization,” “human resources,” and ” governance ,” but he claims that “The Universal AI Platform will be the solution to overcome these barriers.” The Universal AI Platform enables AI development and operational orchestration. In AI development, we provide elements such as generative AI and agents, machine learning, analytics and insights , and data preparation for AI.
In generative AI and agents, we provide the necessary functions from a secure LLM gateway to development and evaluation governance to expand use beyond the PoC stage to enterprise grade. In machine learning, we cover everything from guided AutoML functions to cutting-edge methods, allowing rapid construction and evaluation of machine learning models while maintaining explainability. In the area of analytics and insights, we will evolve BI and analytics to provide an environment where all customers can make decisions based on reliable data. In addition, as data preparation for AI, data connection, cleansing, analysis, modeling, and deployment can be performed in a single environment. Mr. Sato said, “In this way, Dataiku can balance development efficiency, quality, and explainability, and everything necessary for AI development can be completed with a single product. Japanese customers have praised us for being able to realize product blueprints on a single platform.” The operational orchestration area, another element of the Universal AI Platform, provides functionality for safely and securely managing large-scale AI, taking into account AI governance, AI engineering ops, and the AI ecosystem . AI governance involves centrally applying AI governance standards to all data work and managing the AI portfolio .Maintaining visibility into the data pipeline reduces risk. AI Engineering Ops enables automation of data pipelines, management of production environments for models and agents, and operational integration of projects. As an AI ecosystem, it promotes integration while maintaining flexibility for the introduction of existing infrastructure and new tools, and avoids vendor lock-in.
Also Read: BizTech’s AI market partners with robotics developer Inaho
These operational orchestration functions are designed to “adapt incrementally to the organization’s AI maturity, allowing for steady development from the current situation,” Sato explains. The Universal AI Platform’s main feature is its neutrality. “In terms of technology, it eliminates technical constraints while incorporating the diversity of clouds, data platforms, and AI services. It is also neutral to users, and supports a wide range of users, from no-code interfaces for business users to advanced development environments for data scientists,” Sato says. On top of that, Sato says about AI agents, “They have the potential to solve corporate problems and promote business transformation.” “Dataiku aims to further expand the Universal AI Platform to support the efficient and safe deployment of AI agents and create business value,” said Sato. According to Sato, the possibilities that AI agents bring are cost reduction through process automation, efficiency through employee support, optimization through corporate intelligence, and increased revenue through new services and new business models.
However, Sato said that there are challenges to deploying AI agents. These include limited agent functionality due to lack of collaboration between business departments and data teams, inaccuracy of data, lack of permanent connection to data sources, and even ineffective operation of agents due to lack of testing in production environments. In response to these challenges, “Dataiku strengthens enterprise orchestration and continuously optimizes it. It provides a fully managed development environment and ensures that agents created by anyone are properly governed,” said Sato. As part of this, Dataiku offers “AI Agent with Dataiku,” reducing the complexity of AI agents and establishing a mechanism for their effective use. This tool includes components such as “Visual & Code Agent” that integrates no-code and full-code development, “LLM Mesh” that provides access management for various models, “Agent Connect” that centrally manages agents, “Trace Explorer” that visualizes agent operations, and “Quality & Cost Guard” that manages quality and costs.
Through these functions, “it promotes management of reusable assets and knowledge sharing between departments, and creates an environment in which AI agents can be used efficiently throughout the company,” Sato explains. In addition, he says that control functions are important to maximize the value of agents, and that “governance allows for continuous optimization of agents and the company-wide deployment of AI assets developed by various people, including knowledge workers.” Sato lists Dataiku‘s differentiating factors as its ability to enable enterprise orchestration while maintaining neutrality, its ability to achieve continuous optimization, and its ability to take central governance. On top of that, he says, “Dataiku aims to effectively utilize data within a company as an AI agent and incorporate it into business processes.” For the Japanese market, Sato says, “It is important to gradually improve the maturity of AI.” To achieve this, it is necessary to start with data integration and visualization, then move on to predictive analysis and ML implementation, and the integrated use of generative AI, before moving on to the practical application of AI agents. “Ultimately, we need to extract definite value from data and build business models through gradual introduction in phases, without turning AI agents into black boxes. This approach will realize the full-scale use of AI to create value from data, rather than simply replacing personnel or RPA ,” said Sato.
SOURCE: Yahoo
