While many organizations are exploring the potential of AI, a significant gap exists between initial experimentation and full-scale deployment. The latest release of Cognite Atlas AI™ bridges this divide, focusing on accelerating the adoption of production-ready AI agents to unlock business impact at scale. This release empowers organizations to implement AI solutions faster and more confidently through key product enhancements.
Cognite Atlas AI™ is a low-code industrial AI agent workbench that powers agents with real-time, AI-ready operational technology (OT), information technology (IT), and engineering data. Integrated with the AI-ready data foundation of Cognite Data Fusion®, industrial organizations can use AI agents to improve efficiency, reduce costs, and elevate productivity. The core components of Cognite Atlas AI include the low-code agent builder, preconfigured agent templates, purpose-built industrial tools, a curated LLM library to benchmark the best model for specific tasks, secure data and user governance, and integrations through the Agent API.
Anchoring agents in industrial reality
Unlike general-purpose AI, industrial agents must operate within a complex, real-world environment. Cognite Data Fusion and its powerful industrial knowledge graph are key to achieving this. This allows Cognite Atlas AI to build true industrial agents, not just back-office tools. While most AI applications are confined to analyzing spreadsheets and reports, our agents are anchored in the physical reality of industrial operations. They don’t just read data; they understand the relationships between assets, systems, and processes.
For example, our Root Cause Analysis agent doesn't just look at a single data point; it navigates the knowledge graph to trace a failure from a specific sensor reading to the related pump, its maintenance history, and the surrounding operational context. This deep, contextual understanding allows our agents to provide insights that are not only accurate but also actionable in a complex industrial setting.
Here is a closer look at the key enhancements in this product release that accelerate the scaling of agentic AI:
Smarter knowledge graph queries
With the new Agentic Knowledge Graph Queries, agents can perform multi-step queries to provide more accurate and reliable answers. By tapping into the real-time industrial knowledge graph in Cognite Data Fusion®, these agents can now handle more nuanced industrial questions from engineers and other subject matter experts. For example, agents now have access to more information in context, such as edge properties in a data model or diagram annotations in a P&ID. These enhancements provide deeper context within complex datasets, allow experts to access richer insights, and increase the scope of AI agent actions.
Accelerating time-to-value with agent templates
To overcome the struggle of getting started with AI agents for core operations, Cognite Atlas AI™ offers preconfigured agent templates for common use cases. Current Cognite customers can use templates to deploy agents into production in a matter of minutes. For new customers, agents can be deployed in as little as 3-4 weeks. This user-friendly, low-code approach, combined with purpose-built industrial tooling, lowers the barrier for domain experts to build their own agents. Examples of AI agents solving common use cases to deliver value for industrial organizations include:
- Cognite Data Scout agent: Retrieve industrial data across different data types using natural language.
- Time Series Trender agent: Get on-demand statistics and perform analysis to optimize your operations.
- Operations Summarizer agent: Create a unified view into operations across production lines and units from operator rounds, maintenance logs, operator reports, and shift notes.
- Work Package Generator agent: Prepare all industrial data needed for maintenance jobs through a single prompt.
- Work Order Classifier: Automate work order classification to find maintenance patterns and improve data accuracy.
- Document Extractor agent: Unlock critical information buried in documents to enable safer operations, accelerate engineering audits, and future-proof knowledge transfer
- Root cause analysis agent: Receive guided support for troubleshooting steps during critical equipment failure.
Seamless agent interactions
The industrial end-user experience has evolved with AI agents seamlessly embedded across Cognite products, enabling agents to accelerate industrial user workflows and data management efforts. The new cross-agent chat enables users to interact with multiple agents across applications. For example, users can work alongside a root cause analysis agent in Industrial Canvas, a digital collaboration environment, and continue their analysis in Charts, a time series analytics application, without losing context. A refreshed user interface also delivers smoother, more consistent interactions, fostering the trust and transparency needed to become integral to daily workflows.
Document parser enhancements
The Document Extractor agent is more innovative and flexible, unlocking critical information buried in documents. With powerful new features like LLM vision capabilities, setting custom parsing instructions, selecting page ranges, and extracting multiple instances, subject matter experts have more control and better performance, enabling them to retrieve valuable data previously trapped in files easily. For example, one of Cognite’s global component manufacturer customers used the Document Extractor agent to save four months of manual effort by extracting critical information from 10,000 documents covering 3,500 parts.
Secure & governed agent operations
Security and governance are paramount in industrial operations. Cognite Atlas AI™ is built with enterprise-grade security, auditing, and compliance features, employing the robust data foundation of Cognite Data Fusion®. Administrators can define who can use, create, or deploy agents with user permissions. Additionally, there is now an extra layer of protection for high-impact actions. Tools like Python execution and REST API calls now require explicit confirmation to run, preventing accidental execution, strengthening compliance and security, and ensuring proprietary data remains protected.
Accelerating widespread agentic AI adoption
The latest release of Cognite Atlas AI™ is a direct response to the industry's need for scalable, production-ready AI solutions. By integrating directly with Cognite Data Fusion®, Cognite Atlas AI™ can access a trusted, scalable, and AI-ready foundation for industrial data. This unique combination is the fastest way to realize transformation initiatives across operations, helping optimize production, reduce downtime, and increase operational efficiency with industrial AI at scale.
By providing purpose-built industrial tools and preconfigured agent templates, Cognite Atlas AI™ democratizes AI for operations. This release lowers the barriers for both digital and operations teams to build and iterate on AI agents without needing to write a single line of code. With Cognite Atlas AI, organizations can quickly incorporate agents into operational workflows across their enterprise, ultimately empowering a future where human insight and autonomous agents work together to increase production, eliminate unplanned downtime, and reduce waste.