At Impact 2025, Cognite committed to unlocking $100 billion of customer value by 2035. The December 2025 product release is the next significant step toward this goal, focused on creating simplified and intuitive experiences for industrial users, increasing the trust and reliability of AI agents, and enhancing scalability across the industrial knowledge graph. Realizing $100 billion of value requires that both our customers and partners can build and deploy trusted AI agents at scale, automating previously complex and manual tasks and streamlining operational workflows. Cognite’s December product release advances this vision with updates across the industrial knowledge graph, AI agents, and industrial experiences.
Industrial Knowledge Graph: Scale Industrial AI with a trusted foundation
Cognite Data Fusion®, the leading industrial data foundation that makes data AI-ready, has proven its scalability with over 76 trillion time series data points, 260 million files, and 1.5 million field-captured insights contextualized and available for consumption by industrial users and AI agents. In the December release, the enterprise-scale effort continues with the general availability of the Records API and semantic modeling with NEAT.
Enterprise Scalability of Records
Industrial operations generate billions of log entries, events, and historical records each year. Storing this high-volume data as graph nodes can slow down queries. The new Records API unlocks the next level of scalability by enabling the storage of billions of structured records without compromising performance. Records seamlessly integrate with Data Modeling, allowing for direct relations to nodes in the Knowledge Graph, providing a robust, scalable structure for both live data and long-term archives. Information like machine states, alarms, and events are available in context to be accessed by industrial users, developers, and AI agents through Cognite Atlas AI.
Cognite NEAT: Empowering developers and domain experts
A reliable, scalable, and usable data model is the bedrock of the Knowledge Graph in Cognite Data Fusion. The Cognite Knowledge Graph Transformer (NEAT) is an open-source, developer-friendly tool that serves as a powerful bridge between general semantic web modeling principles and the core capabilities of Cognite Data Fusion, making it accessible for both domain experts and developers. Incorporating years of product and customer expertise, NEAT is the quickest way to operationalize AI-ready data models in Cognite Data Fusion, reducing the time to build data models from weeks to hours.
In this release, NEAT becomes an officially supported developer tool, marking a significant milestone in Cognite's commitment to provide the most developer-friendly industrial AI and data platform. Now developers can:
- Ensure data models are production-ready via several interfaces with pre-deployment analysis and configurable governance profiles
- Work in spreadsheets, notebooks or in scripts with readable, object-oriented Python code
- Validate your data model and fix issues quickly with interactive error feedback
As an open-source tool, NEAT is fundamental to the open ecosystem surrounding Cognite, providing developers with the resources needed to rapidly build and scale tailored solutions using semantic data models.
AI Agents: Increasing consistency and reliability of AI agents
Building trust and ensuring the continued performance of AI agents remains critical for enterprise adoption. Cognite Atlas AI™, the only low-code industrial AI agent workbench to power agents with real-time, AI-ready operational technology (OT), information technology (IT), and engineering data, continues to see significant market momentum and onboard new customers every week. This December release simplifies the effort to ensure consistency and reliability of AI agents.
Agent performance evaluation
Users deploying agents need visibility into how agents are performing over time to build trust, improve quality, and troubleshoot failures. The new Evaluate Agents feature enables users to create and manage test sets with prompts, ensuring expected responses from agents. This ability to run evaluations, view pass/fail results, and inspect individual agent responses enables data-driven confidence in agent reliability. This feature also accelerates agent deployment by better supporting User Acceptance Testing (UAT) and delivery handoff, while also establishing a foundation for long-term quality tracking and regression checks.
Consistency in knowledge graph queries
AI agents built with Cognite Atlas AI are grounded in an industrial organization’s industrial knowledge graph. These agents have access to purpose-built industrial tools for specific tasks, including the ability to query the industrial knowledge graph for specific data types such as time series, assets, and work orders. To further increase the reliability of agent behavior and retrieval consistency, users can now define static filters and properties for the Query Knowledge Graph tool. This feature defines how the agent queries data, ensuring it retrieves consistent and relevant results by configuring which filters and properties the tool uses, increasing trust and transparency in agents.
Industrial Experiences: Intuitive tools for industrial users
To realize value and further automate operational workflow, industrial users deserve a seamless and intuitive experience. This release includes enhancements across search and industrial applications, such as Industrial Canvas.
Enhanced search, 3D, and Industrial Canvas Experience
In Search, first-time users now have a more guided experience. New features also allow users to view more instances at once, switch between list and grid views, and scroll within tables while keeping column headers, filters, and view options visible. Search also supports additional language structures to provide more relevant search results when searching in languages such as Japanese.
3D continues to advance with more user-friendly features. Users can now perform one-click diameter measurements for pipes, vessels, and tanks, eliminating laborious field work. Customizable settings per scene allow users to define the default model visibility, rendering quality, and Point Cloud settings. This reduces friction for regular users and improves the first-time experience.
Usability improvements in Industrial Canvas make complex data flows easier to understand. When multiple document annotations point to the same destination, their connection lines are now merged into a single line and route around other documents with 90-degree angles instead of overlaying them. This improved display logic makes the canvas more readable, particularly when performing activities such as troubleshooting, planning, or root cause analysis across complex diagrams and processes.
The December 2025 product release marks the next step in Cognite's journey toward unlocking $100 billion of customer value by 2035. These advancements across the Industrial Knowledge Graph, AI Agents, and Industrial Experiences are critical for industrial organizations to successfully integrate trusted AI agents into their core operational workflows. To learn more about how Cognite is unlocking industrial AI at scale, see the product keynote from Cognite Impact.