Cognite’s latest product release furthers its leadership in industrial AI and DataOps by increasing the usability and flexibility of AI agents for AI-powered insights.
The newest release of Cognite’s industrial agent workbench, Cognite Atlas AI™, allows users to easily customize AI agents and increases the accessibility of real-time data with REST API support. Additionally, the latest release of Cognite Data Fusion® enhances collaboration amongst industrial users for troubleshooting and root cause analysis and accelerates production optimization workflows with Petex simulator integration.
Transform traditional industrial workflows with AI-powered insight

With the release of the REST API agent tool for Cognite Atlas AI, AI agents now have access to the most relevant, up-to-date information in Cognite Data Fusion’s industrial data foundation. Previously, Atlas AI relied upon pre-built configurations to access data in Cognite Data Fusion. Now, AI agents can use the API agent tool to fully customize access to real-time data in Cognite Data Fusion, increasing the breadth of workflows. The combination of AI agents and native Cognite API integration expands the scope of AI-powered capabilities, including time series anomaly detection, real-time troubleshooting and analysis, and work order creation. Aligning with Cognite’s commitment to openness, this native integration is a precursor to an external API integration, extending Atlas AI agents into workflows of 3rd party applications.
To simplify the efforts to build and tailor AI agents, Cognite Atlas AI has now embedded a Python script editor, enhancing agent functionality without external dependencies. This streamlines workflow automation and empowers builders to customize AI agents in a unified experience.
As part of this product release, Cognite has also released a second edition of the Cognite Atlas AI LLM & SLM Benchmark Report for Industrial Agents. This report, designed to help industrial organizations identify which language model should be used for specific industrial tasks, expands on its first release by introducing Document Question Answering (Document QA) to complement the previously covered natural language search (now Natural Language Query). Document QA addresses another critical capability: enabling AI agents to analyze and extract insights from industrial documents, such as manuals, maintenance logs, reports, and compliance records. Together with natural language search, this forms a holistic foundation for building reliable industrial AI agents.
Real-time Production Optimization with Petex Simulator Integration

Upstream energy organizations have made significant investments representing their operations in advanced modelling and simulation software for production optimization. Simulators often require time-consuming manual processes to batch data for updated insights and recommendations. With this release, Cognite Data Fusion now supports connectors native integration for Petex PROSPER and Petex GAP, unlocking real-time optimization of wells and networks and increased operational efficiency. The Petex connectors automate the data connectivity of contextualized industrial data from Cognite Data Fusion to Petex simulators, eliminating the manual efforts to maintain simulation models and access insights.
Streamlining collaboration for troubleshooting and root cause analysis
Industrial organizations such as Nova Chemicals are using Cognite Data Fusion’s Industrial Canvas to analyze operations, troubleshoot issues, and perform root cause analysis (RCA) on unplanned events. Cognite’s industrial canvas provides an unbounded workspace to engage with all contextualized information, such as time series, work orders, documents, drawings, 3D, and more, in one unified environment.

Industrial Canvas already incorporates AI-powered document summarization and questions, along with a powerful toolbox for users to easily find relevant information and add situational context when analyzing operations. As part of this release, the ability to create tables has been added to the toolbox to complement existing toolbox functionality, such as notes, shapes, lines, and more. Cognite continues to invest in AI and toolbox capabilities within Industrial Canvas to help industrial users streamline data-heavy workflows such as RCA, work package creation for turnaround planning, troubleshooting, and more.

Another challenge organizations face with troubleshooting and RCA workflows is retaining knowledge from previous assessments and recalling the corrective actions implemented in a previous analysis. To increase the visibility and collaboration around historical knowledge, Industrial Canvas now includes labels to organize and search for relevant information. By applying one or several labels to your canvas, users can search for analyses by specific equipment types, functional areas, or process types, allowing knowledge and corrective actions to be shared across the organization.
To experience the power of our latest release, join us at Impact 2025 – the premier industrial data and AI conference for users, by users. See firsthand how we're shaping the future of industry.