<img height="1" width="1" style="display:none;" alt="" src="https://px.ads.linkedin.com/collect/?pid=1205042&amp;fmt=gif">

Cognite AI: An Open Letter to Customers


Share on:

This is the moment that Cognite was built for.

Industrial reality is complicated. 

I know it well. Over 60 years ago my father was a foreman in a chemical plant in Baton Rouge, Louisiana, and eventually a production manager in an aluminum company in the high hills of India. Our home had early morning echoes of his update calls: animated seeking of accurate production data, reason for downtime, resource overrun, unavailability of equipment, and yes, at times a somber silence as news of a safety incident that impacted a worker reverberated through our house. Even in those young years, I did not need to be told or learn at school that parsing through data and then acting upon it was the nature of manufacturing. That aspect was an omnipresent airwave in my childhood home. 

That aspect is a stark realization in my professional years serving industry, that actionable data can make companies, communities, and countries. Software done well can make communities safe, healthy, livable. Recent developments in software provide inroads for progress in industrialization that eclipse the progress made during the industrial revolution. 

Cognite has worked with over 100 global energy, utility, and manufacturing companies to make data available from cross-data sources to operators. Accurate, timely, contextualized data. Be it in in-field applications used by industrial users in a large chemical plant in Texas or a global optimizer in an energy company in the middle east—the task of unifying and contextualizing operations, engineering, and enterprise resource planning to power modern-day manufacturing is an urgent want of our customers. Cognite is delivering upon it. 

Artificial Intelligence (AI) and the availability of Large Language Models (LLMs) is the next horizon for democratizing data. However, in an industrial environment "almost" accurate data, "probabilistic" data, parsing through garbage called "hallucinations" just does not work. In addition, the thought of putting proprietary enterprise data or its leakage in the public domain is nothing short of corporate blasphemy. 

Today we announced Cognite AI, a comprehensive suite of Generative AI capabilities within Cognite’s core Industrial DataOps platform, Cognite Data Fusion®. Cognite AI expedites the application of modern software technology in industrial digitalization and accelerates time-to-value from Generative AI. It informs the industrial user, empowers the industrial operator, and enables industrial facilities to be efficient, sustainable. Cognite Data Fusion® with Cognite AI will be used across oil drilling and wind farms, refineries and utility grids, mines, and factories for revenue generation, cost reduction, and risk avoidance. 

Key to Cognite AI is an innovative architecture that unifies Generative AI with Cognite Data Fusion's specific Data Modeling and Retrieval-Augmented Generation (RAG) capabilities. This is how we eliminate hallucinations, mitigate data leakage, and enable trust and access control. This is how we make Generative AI work for industry. 

I. The Context Behind Contextualization

Industrial data has been siloed, inaccessible, and too static to be useful for meaningful decision-making. It provides a reasonable historical perspective, a system of record at a nodal level. Traditional approaches to industrial data provide insights from a single source of data. Traditional approaches are complex, drowning in technology debt, part computing on servers and part shifting to the cloud—a Sphinx at best. Traditional industrial applications run high maintenance costs throughout the life of the system. Custom code stitching creates bespoke applications that strain on scale. 

The Industrial DataOps foundation provided by Cognite Data Fusion® solves these limitations. The power of Generative AI coupled with data modeling enables us to derive hidden relationships from cross-data sources. Cognite AI makes these cross-data-source insights available for asset performance management practitioners and other users in an easy and friendly way. 

II. Seize the Moment: Generative AI

With Generative AI and LLMs, a query formed as a natural language sentence (a prompt) is replacing certain software applications built for data exploration, and bypassing the traditional application development cycle. This is significantly reducing the time and cost needed to address a business user’s specific data needs. As a result, Cognite AI meaningfully expands the number of users who can access the data they need and lowers the required skills to do so. 

Cognite AI has integrated Generative AI capabilities, enhancing Machine Learning capabilities within Cognite Data Fusion® to make it easier to ingest structured and unstructured data, and to automatically infer physical and virtual relationships from various data sources in the industrial environment.

Cognite AI enhances data exploration with the integration of LLMs, in particular, GPT4 through our close partnerships. 

The software we have built enables our industrial customers to onboard their operational data sources including but not limited to time-series, events, asset information, documents, piping and instrumentation diagrams, audio, video and enterprise resource planning data. Cognite AI automatically detects and maps relationships from physical assets and environments to the data generated by them. This is contextualization on steroids. We have made contextualized and enriched data available and consumable through a set of 1st party (developed by Cognite) and 3rd party (developed by Cognite partners and customers) applications. 

Here are a few examples of Cognite AI use cases that are available for customers.

III. Why AI With Cognite?

LLMs and Generative AI platforms like ChatGPT add new valuable capabilities to software applications. At the same time, since they are trained on public internet data, they tend to generate false information when the query depends on private data that is not available on the public internet. This is called hallucination

An LLM like ChatGPT is just that—a large language model and not a data or an image model. It is a model that is trained for language, and it can output plausible words in a sentence based on probabilities given an input. 

That mode is not conducive for industrial usage—we don’t want data in the output to be probabilistic. Industrial users need facts to run their mission-critical environment. Cognite AI includes flexible data modeling with all the data in it, that is queryable with a programming language (API) that can deliver these facts. Cognite AI uses the large language model to translate the user query to our programmatic query, that is also a language, and this translation is exactly what LLMs were built for. 

With this approach, we get the best of both worlds:

  • The large language model translates human language to programmable language.
  • Every single user in your organization can use it because it only requires them to write what they want in human language.
  • The flexible data model in Cognite Data Fusion® delivers the data with: no hallucinations; no data leakage; full access control, data lineage, and high data quality. 

Thus: Cognite AI, through an innovative architecture, couples Generative AI and its LLMs with Cognite Data Fusion’s specific Data Modeling and Retrieval-Augmented Generation (RAG) capabilities. Combined, Cognite AI can uniquely augment the ability of generic-purpose LLMs to retrieve data from customers’ private sources to generate more purposeful, accurate, and sophisticated outputs based on customers’ private industrial data, within customers’ secure and protected SaaS tenant.

IV. The Industrial User Deserves Cognite Data Fusion® With Cognite AI

Cognite Data Fusion® is the only platform in the industry that makes cross-data source insights available to users in operations as well as to data scientists, data engineers, and software engineers to build applications for field workers and asset operators. 

Our values stay true to our genesis in Norway. We aspire and strive to "make an impact" via our software—on users, operations, enterprises, companies, and communities. As a result, Cognite is committed to sharing our platform and approaches with our stakeholders in industry, technology, academia, and at think tanks. 

V. The Pressing Question

The memories of my father working the factory floor, immensely personal to me, are memorialized. Today's pressing question is, however, what will your role be in this seismic new AI frontier? This unprecedented opportunity to harness technology to transform industry will demand you choose a lane. Will you take up the mantle and co-innovate scalable solutions that define the future of companies, communities, and countries? Each day Cognite further connects the dots of our work to the causes of the energy transition, carbon capture, and a more sustainable and efficient future of manufacturing operations. We look forward to bringing Generative AI to your industrial digitalization journey and co-innovating together.

This is our customers' moment. This is our moment. 

Girish Rishi

CEO, Cognite



Share on: