Dataiku Kiji is advancing enterprise AI explainability as organizations scale AI agents across critical workflows. Performance alone is no longer sufficient. Enterprises now require visibility, validation, and governance of how AI systems generate outputs, especially in regulated environments.

Through its 575 Lab open-source office, Dataiku has introduced Kiji Inspector™, an open-source explainability framework designed for enterprise AI agents. Notably, the first model family supported by the framework is NVIDIA’s Nemotron open models.

As enterprises increasingly adopt sovereign AI and build internal infrastructure, the combination of NVIDIA’s Nemotron models and Dataiku’s Kiji Inspector enhances visibility into AI-driven decisions. As a result, organizations can better understand how outputs are generated. The framework directly addresses the long-standing black-box problem in enterprise AI.

At its core, Kiji Inspector™ uses a Sparse Autoencoder. It analyzes model behavior at the moment an AI agent commits to a tool. Then, it identifies key signals behind that decision. These signals are translated into clear, traceable explanations. Importantly, this process does not slow system performance.

According to Hannes Hapke, Director of 575 Lab at Dataiku, enterprises are embedding AI agents into decisions that impact revenue, safety, compliance, and customer trust. However, many still lack structural visibility into AI reasoning. He stated that without explainability, scaling AI also scales uncertainty. He added that integrating Kiji Inspector with NVIDIA Nemotron models enables organizations to refine explainability before risks emerge, especially as agentic systems evolve into trusted infrastructure.

Meanwhile, the release builds on the ongoing collaboration between Dataiku and NVIDIA to deliver production-grade generative and agentic AI. NVIDIA’s Nemotron open models offer advanced capabilities required for enterprise AI systems. In parallel, Dataiku provides scalable orchestration by connecting data platforms, enterprise applications, and AI services within a governed framework.

Amanda Saunders, Director of Generative AI at NVIDIA, emphasized that scaling autonomous AI agents requires trust built on transparency and accountability. She noted that open models like Nemotron provide visibility into system operations. Furthermore, combining these models with Kiji Inspector allows users to understand what drives large language model decisions.

In addition, Dataiku’s focus on explainable AI aligns with the needs of its 750+ enterprise customers. Many operate in complex and regulated industries. For instance, SLB, a global energy technology company, highlighted the importance of transparency as it expands AI adoption.

Sampath Reddy, Global Innovation Manager – Data & AI at SLB, stated that AI delivers value in energy operations when engineers can understand and trust its decisions. He added that validated workflows and strong governance enable teams to deploy and scale agentic AI confidently in real-world environments.

Overall, by extending Kiji Inspector to NVIDIA Nemotron models, Dataiku enables enterprises to leverage high-performance open-source AI without sacrificing transparency. As AI agents become more autonomous, explainable reasoning is expected to become essential for long-term success, regulatory readiness, and competitive advantage. Therefore, Dataiku Kiji is positioned to play a key role in building trustworthy and scalable enterprise AI systems.

Related post

View all