French enterprise AI company Dataiku presented "AI internalization" as the key to corporations' competitiveness in the AI era and unveiled the AI agent "CoBuild."
Dataiku held a press briefing for "Dataiku Summit Seoul 2026" on the 24th at Episode Gangnam 262 in Seocho-gu, Seoul, and introduced strategies to boost corporations' results in adopting AI. Andrew Boyd, senior vice president and general manager for Asia-Pacific and Japan (APJ) at Dataiku, and Jean-Guillaume Apper, director of product management in the institutional sector, attended the briefing.
Vice President Andrew Boyd said, "According to a McKinsey survey last year, 88% of corporations worldwide are using AI, but only 6% have experienced tangible value from AI," and added, "Many corporations remain in the pilot stage; AI expenditure is increasing, but they are struggling to generate results."
He defined this as the "AI Success Divide" and stressed that the success or failure of AI depends on how effectively it is internalized across the organization. Simply having data, platforms, and various AI tools cannot guarantee results; AI must be applied to actual work processes and decision-making to lead to productivity gains.
Boyd said, "The value of AI is revealed not by the technology itself but when it operates within the real organization," and added, "For AI to be properly internalized, business users must be able to understand it easily, risk management organizations must be able to control it, and the IT department must be able to operate it reliably. The Dataiku platform is designed to meet these conditions simultaneously."
Next, Jean-Guillaume Apper, director in the institutional sector for product management, introduced the AI agent CoBuild, which was released on the 18th. CoBuild is a platform that automatically converts business goals entered in natural language by frontline users into operable AI projects.
When a user describes a goal, CoBuild identifies relevant data and designs the workflow, then automatically generates data pipelines, Machine Learning (ML) models, AI agents, and applications without separate coding. The generated outputs are provided as a visual flow so that frontline departments and the IT department can jointly review, modify, and approve them.
In particular, it is designed so that all development processes and outputs can be checked visually, allowing governance and compliance to be naturally reflected from the initial planning stage. It also supports a variety of large language models (LLM) and data environments, including Snowflake, OpenAI, Anthropic, AWS Bedrock, and Google Gemini, enabling corporations to choose the models and platforms they want without being locked into a specific vendor.
Apper said, "While existing Generative AI focused on quickly producing outputs, CoBuild visually shows even the process by which outputs are created," and added, "Frontline users can understand and directly validate the structure and flow of AI projects, and the IT department can manage them while maintaining governance and compliance."
Kim Jong-deok, head of Dataiku Korea, said, "In the era of agentic AI, what matters more than the performance of individual models is the system that can safely scale and control them," and added, "Dataiku will continue to support domestic clients so they can develop AI not as one-off experiments but into companywide operating systems."