From Aging to Agent-ing with Cognite’s Atlas AI

You may have heard whispers (or even shouts) about industrial agents being the next big thing within generative AI for industry. And rightfully so! Industrial agents can perform specific tasks, understand industry and company-specific processes, and help workers make decisions by analyzing historical patterns and real-time data. Providing insights and solutions in a human-like manner, industrial agents move us closer to bringing Iron Man’s Jarvis to life.

Since the onslaught of AI last year, tech innovators have learned a lot. A main takeaway for Cognite is that a one-chatbot-to-rule-them-all approach does not work. Heavy-asset industry needs AI that can complete specific and complex tasks. These specific tasks therefore need tailored and relevant solutions, not just an AI jack-of-all-trades.

Industrial agents: The Digital Ava

Imagine this scenario.

You’re a new worker at a big factory. The most experienced worker there, Ava, is showing you the ropes. It seems Ava has a solution to every problem on the factory floor stored in her head. During your first week, you go to Ava with all the questions you might have. The next week, Ava is gone. She retired. The only thing she left behind was a massive handover document with links, links within links, and more links. As in most cases, the factory you work for does not have basic DataOps and data management. This means you have to spend hours digging through links and data sources while mapping different source systems to find anything close to data-driven insights.

Let's say the factory is further in its digital journey and has made steps toward solving the data problem. The organization has already connected data sources, contextualized data, and improved data quality to build a knowledge graph. With simple access to this contextualized data through the knowledge graph it is more likely that you can gather the data needed to analyze questions such as: “What is the recommended interval for performing vibration analysis on a pump to prevent potential failure?” If only you had a digital Ava who could answer your industry-specific questions….

Heavy-asset industries are all trying to tackle the big elephant on site; the aging workforce. Left unaddressed, it poses a serious threat. Within manufacturing:

• Brain drain–the loss of institutional knowledge from experienced workers retiring–costs businesses $47 million per year due to time waste, missed opportunities, frustration, and delayed projects, according to the Association of Equipment Manufacturers.
• 97% of firms expressed concern about brain drain due to the retiring workforce, according to the Manufacturing Institute.

As professionals retire, they take years of knowledge and experience with them, and handover documents provide little value compared to the deep technical and institutional knowledge of veteran professionals. As employees enter heavy-asset industries, a massive skill gap becomes evident, where new workers need to catch up to those who have decades of experience. It can take years before new workers gain the know-how necessary to match their more experienced coworkers, and with an aging workforce, this skill gap will only continue to grow.

So how does Cognite’s Atlas AI address this?

Companies need a Workbench for their Industrial Agents

Cognite’s Atlas AI provides a low-code industrial agent builder that enables organizations to use generative AI to address domain-specific challenges with a deep understanding of industry context, terminology, and workflows. These agents leverage advanced technologies to provide expert guidance, and offer highly relevant insights in support of specific tasks.

Atlas AI is powered by Cognite’s unique combination of industry-tailored AI product capabilities, specialized industrial domain knowledge, and industrial generative AI delivery expertise. As an industry and Gen AI delivery expert, Cognite’s Atlas AI provides everything necessary for asset-heavy organizations to orchestrate industrial agents on top of their private data.

A library of pre-built industrial agents, as well as a low-code agent builder

Pre-built library

Cognite is already working on several agents. Take, for example, an industrial agent for document parsing. This agent is fine-tuned to read technical documentation and unstructured input such as data sheets to find relevant information and fill out specific data in a structured form.

Low-code agent builder

The Cognite Atlas AI agent builder allows users to create agents on their private data to fulfill industry and company-specific needs. The agent builder includes agent templates that users can easily modify to fit their requirements.

With an accessible pre-built library and a tailored low-code builder, users can choose what best fits their needs, requirements, and restrictions.

A Semantic Industrial Knowledge Graph with Context Augmented Generation (CAG)

Semantic Industrial Knowledge Graph

A knowledge graph is a structured representation of information that allows industrial organizations to better understand the relationships and connections within complex datasets. Benefits of using a knowledge graph include:

• Providing a unified view of data, breaking down data silos, and improving understanding of usage and consumption patterns
• Categorizing and managing information using standardized metadata, enhancing the utilization of unstructured data in documents, images, and videos, and turning that data into actionable insights.
• Empowering AI applications to provide high-quality, trusted insights through the contextualized knowledge, rules, and semantics embedded in knowledge graphs.

By breaking down data silos and providing contextualized data, knowledge graphs serve as the foundation Gen AI requires to create actionable insights.

Cognite’s Atlas AI combines two types of knowledge graphs:

• A Semantic Knowledge Graph that captures the relationships between entities in a way that humans and machines understand. It represents a time-saving tool that streamlines manual data collection and integration efforts to bolster decision-making processes.
• An Industrial Knowledge Graph that focuses on the intricacies and specifics of industrial processes, machinery, operations, and related data.

Context Augmented Generation

CAG is the evolution of Retrieval Augmented Generation (RAG). RAG retrieves information from external databases to enhance generative models. CAG does this while integrating content and data from multiple sources, such as real-time data, sensor inputs, user interactions, and historical data. CAG allows AI systems to give and create more complex and content-aware responses.

The combination of a Semantic Industrial Knowledge Graph and CAG ensures that insights created by Industrial Agents are accurate, industry-relevant, and valuable.

Industrial use case-tailored autoLLM capabilities to identify the best Large Language Model, Small Language Model, or Custom Language Model for any given industrial agent

Three types of language models are available for industrial agents: large, small, and custom. The choice of language model depends on the industrial agent's requirements, constraints, and goals. Navigating these choices can be challenging, time-consuming, and requires extensive technical knowledge. To address this, Cognite offers AutoLLM, an automated process that helps users select the most suitable language model for their industrial agent needs.

AutoLLM evaluates the best language model based on a provided use case, performance expectations, resource constraints, and budget. The system uses algorithms to match user requirements with the most appropriate models from the database. AutoLLM then optimizes model performance by automatically tuning the recommended model for a specific task.

Easy deployment of industrial agents into Cognite Data Fusion®, Microsoft Copilot, proprietary applications, or other third-party applications

Industrial agents made with Cognite Atlas AI are built with cross-application functionality in mind. The agents are designed to work seamlessly with industry-specific software, hardware, and operational protocols. This integration ensures minimal disruption and allows for smoother implementation and higher compatibility

Bridging the Skill Gap

These features allow domain experts, not just data scientists and programmers, to build industrial agents that are accessible, industry-relevant, valuable, and highly usable.

But how do these industrial agents address the aging workforce?

Industrial agents built by Cognite’s Atlas AI help bridge the skill gap by empowering new workers with data-driven insights to make better decisions without requiring years of experience. These industrial agents understand industry and company-specific processes, equipment, and best practices, assisting workers in making informed decisions by analyzing data and historical patterns and providing relevant information based on real-time data.

In short, Cognite’s Atlas AI provides the workbench companies need to build their own digital Ava’s; industrial agents that understand company-specific practices and provide industry-relevant insights and solutions.

Agents are already being deployed

You might be thinking: This all sounds good in theory, but how far-fetched are these industrial agents? How many years will I have to wait to deploy something like this?

Cognite has already deployed several agents. Take, for example, an industrial agent for document parsing, built using Cognite Atlas AI. This document parsing AI agent is fine-tuned to read technical documentation and unstructured input such as PDFs and data sheets to find relevant information and fill out specific data in a structured form.

Automating data extraction eases the workload for process engineers. More than this, however, this industrial agent document parser reduces the skill gap left by the aging workforce, as less experienced field workers and maintenance teams have readily available, high-quality information to act upon.

AkerBP has deployed this document parsing AI agent to automate equipment registration as part of their digital transformation journey, which is already forecasted to save AkerBP more than 10,000 engineering hours.

“Aker BP aspires to lead the energy sector in data-driven innovation. Cognite Atlas AI enables us to use AI to enhance decision-making and improve efficiency, like with our joint Document Parser AI Agent,” said Paula Doyle, Chief Digital Officer at Aker BP ASA. “This industrial agent is fine-tuned to understand unstructured technical documentation and Aker BP’s equipment hierarchy to streamline our equipment management process. By implementing the Document Parser AI Agent, we have saved thousands of hours of data-punching and refocused our experts on business problems that really matter to the short- and long-term success of our operation.”

Agents will transform all aspects of industrial operations in the future

Combining domain expertise and contextual, industry-relevant insights, Cognite Atlas AI represents a major step in reducing the skill gap left by the aging workforce. Industrial agents empower new workers, providing them with a digital Ava to help them make smarter, data-driven decisions.

But It does not end there. As tailored, efficient, and practical solutions, Agents do more than just create insights. They can also address unique, industry-specific challenges, improving safety, efficiency, and decision-making in heavy-asset industries. Examples include:

While Cognite’s Atlas AI helps reduce the effects of an aging workforce, it has the potential to do so much more.

Its value is evident. See how you can implement Cognite Atlas AI to improve your industrial operations.