Cognite’s New Interface Gives Industry Its First Real Glimpse Of Generative AI’s Value

Cognite’s New Interface Gives Industry Its First Real Glimpse Of Generative AI’s Value

Sure, we’ve all asked Bard, Bing or ChatGPT to draft us an email, rephrase a paragraph, or write a poem about something inconsequential. However, despite the flashy demos, billions in new funding and millions of social media hustle bros posting about the newest AI tool, generative AI is still very much finding its feet. With the proliferation of publicly available and open-source large language models (LLMs), the business case for consumer applications is increasingly being drowned out by intense competition from anyone with a basic understanding of computers and an evening to read about AutoGPT or LangChain. And, of course, LLMs are superb at helping novices develop functional software from the ground up.

Looking specifically at software serving heavy industries – oil and gas, power utilities, manufacturing, mining, infrastructure – we saw a flurry of press releases effectively providing a “yes, we know!” statement or brief descriptions of intent to integrate generative AI into products. The problem with industrial applications using generative AI is that mistakes can result in millions of dollars in lost revenue, pollute entire ecosystems and even cost lives. Plus, most industries already create more data than they can handle – so much so that Verdantix has two recent reports on the topic: a study of the importance of DataOps platforms and an analysis of 20 vendors offering solutions in this space.

DataOps-focused vendors are uniquely placed to harness the power of AI/ML technologies such as LLMs for asset-heavy industries – taking advantage of industry-specific data ingest, transform, governance and contextualization tools to provide a unique grounding plane for hallucination-prone LLMs. In the first half of this year, two such vendors threw their hats into the ring: C3 AI, with its Generative AI Product Suite, and Palantir, with its Artificial Intelligence Platform (AIP).

On June 15^th, another industrial-focused DataOps vendor, Oslo-based Cognite, formally launched “Cognite AI”, described in its blog as a “comprehensive suite of Generative AI capabilities” within its Cognite Data Fusion platform.

Previous Verdantix consideration positioned Cognite as a vendor more focused on industrial customers with pockets deep enough to support an in-house data science team and lengthy implementation projects with help from third-party consultants. Cognite AI circumvents this criticism by using the natural communication abilities of LLMs to act as a general-purpose low-code interface to its solution’s advanced features – providing far more of the workforce access to valuable insights. Cognite claims to achieve this without sending sensitive customer data to third-party LLM APIs, like OpenAI’s GPT-3.5 and GPT-4.

Capabilities delivered by this new product release include the Cognite Copilot, a chatbot-style interface where questions such as “What information do we have on the heat exchanger 9115?” are answered with an assortment of P&IDs and 3D models, which offer clickable labels to drill down to the raw data. This Copilot also integrates into their new auto-generated and user-configurable GUI, Industrial Canvas. Operating on a “single pane of glass” concept, the value of which has already been demonstrated by firms such as Viewport.ai, this offering boasts the addition of an in-context LLM-based chatbot and agent-like reasoning engine. At time of writing, available LLM-based solutions by C3 AI and Palantir offer a more structured, less configurable chatbot-style interface, with drill-downs requiring an exit to their existing GUI.

Key to Cognite’s product is their Industrial Knowledge Graph, an enterprise-scale, cloud-first graph database architecture – where disparate information is interconnected by semantic relationships. This approach helps data scientists draw insights from sprawling multinational industrial operations – and offers the same benefits to LLM-based reasoning engines.

While it is still early days, the operational value and benefits of a universal human-machine interface enabled by generative AI in industrial software – such as EAM, APM and AIP – is becoming clearer.