Artificial intelligence is at a pivotal point in its evolution, moving into a new era that goes beyond simple pattern recognition to reasoning, and causal AI is at the forefront of this evolution.
Causal AI offers insights not just into what is happening, but why. This leap in decision-making intelligence has the potential to redefine the marketplace as businesses use these tools to enable smarter, more responsive processes. This next phase of AI evolution will shape the future of the AI ecosystem. Causal AI, unlike traditional models that rely on statistical patterns, is designed to provide explanations and reasoning, according to Scott Hebner, principal analyst at theCUBE Research.
“A lot of people talk about generative AI … but as a leader, you also have to be thinking ahead, particularly with AI, which is moving at an even faster pace than previous technological transformations,” Hebner said. “It’s important to take a futuristic view … so I’m doing a series of five papers about the advent of causal AI.”
Hebner spoke with theCUBE Research’s Principal Analyst Rob Strechay, during an AnalystANGLE segment on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed how causal AI will enable a more structured approach, integrating large and small language models to build a more cohesive, responsive AI and machine learning ecosystem.
Understanding cause and effect
As organizations push the boundaries of AI, they are realizing that today’s models — particularly large language models — are effective at identifying patterns and making predictions but fall short in explaining the reasoning behind those predictions. LLMs operate on statistical probabilities, which are useful, but can be limiting in dynamic, ever-changing environments.
“Today’s predictive models and generative AI models that are embodied in the LLMs are pattern recognition machines. They operate on statistical probabilities … in a static world,” Hebner said. “What causal AI will tell you is how those statistical probabilities change when the world around you changes.”
Causal AI begins with agentic AI, which brings together AI agents in an ecosystem of AI large language models and domain-specific small language models to understand cause-and-effect relationships, a critical factor in helping humans problem-solve and make better decisions, as discussed in the article “The Causal AI Marketplace,” authored by Hebner.
Organizations are constantly in flux, and for AI to truly understand how the business operates, it needs to be able to understand cause and effect, according to Hebner. Why? Because in business everything is a cause and everything is an effect — and AI needs to keep up with that reality.
“Causal AI is all about helping people understand how the business operates. Then from there, it supports a dynamic world of change,” Hebner said. “It’s going to allow those statistical models, probabilities that traditional AI and machine learning operates upon, to adapt.”
The ability to simulate and test what-if scenarios based on the model is another benefit of causal AI. It offers businesses the flexibility to prescriptively model best-case outcomes impacting scenarios around profitability, customer retention and revenue.
“Today’s models are pretty good at predicting what you should do [and] forecast, and they generate the what, but they can’t tell you how it did it. And it certainly can’t tell you why this is the best answer,” Hebner said. “Causal AI is going to start to incrementally allow that explainability to be infused into these models, not only descriptively and predictively, but … prescriptively.”
The role of specialized AI models and agentive AI
While LLMs provide a general-purpose framework, small language models are designed for targeted tasks, allowing businesses to optimize AI for specific needs. These models ensure high data protection and specialized application.
“You need small language models that are specialized, secure and sovereign, that understand each of the domains within a business,” Hebner said.
He envisions a network of SLMs and LLMs where AI agents collaborate and contribute specific expertise. “The whole thing is going to come together in an architectural approach, and that’s going to represent the future,” he added.
This architectural approach will allow AI systems to interact with each other more effectively. LLMs will provide general knowledge, while SLMs focus on specific domains, creating a seamless flow of information, Hebner explained.
“We’re moving toward an ecosystem where AI agents teach each other, learn from each other and become smarter and smarter,” he added. “It’s going to be an architectural approach where agents work collaboratively, and that’s going to be key to the future.”
The case for causal AI
Causal AI isn’t just a concept on the horizon — it’s already gaining traction in industries that require a deeper level of decision intelligence. A recent Databricks Inc. and Dataiku Inc. survey of 400 AI professionals shows that over half of them are already using or experimenting with causal AI, which is expected to be one of the most adopted AI technologies in the coming year, according to Hebner.
“The number one technology that’s not being used today, but they plan to use over the next year, is causal AI,” Hebner said. “[Customers] want to build higher [return on investment] use cases, which require … reasoning, decision intelligence problem-solving and explainability.”
As the demand for more explainable and adaptable AI grows, causal AI will likely play an increasingly critical role in how businesses leverage artificial intelligence for better decision-making. The future of AI, according to Hebner, will be shaped by its ability to understand cause and effect. This shift could redefine how companies approach problem-solving and decision-making in an increasingly dynamic marketplace.
“Gen AI is the big thing today. It wasn’t five years ago. I think over time, causal AI and the notion of why things happen and [questions such as], ‘What can I do to improve things?’ will become a bigger and bigger part of the mix here,” Hebner said.
Here’s the CUBE’s complete AnalystANGLE segment with Scott Hebner: