Agentic AI: Exploring its scope, applicable use cases and current state
Evolution is the lifeblood of technology. Today's digital landscape is flooded with artificial intelligence models serving many use cases. For the most part, however, they share common logical and architectural underpinnings. Agentic AI is being presented as a new way to approach the tech, giving it greater scope, agency and autonomy in executing designated tasks.
So, what is agentic AI? At which potential and existing use cases does it excel? And most importantly, is it the natural evolution of AI as we know it today?
"We've been building AI systems using very similar approaches, architectural patterns like this for years and years and years," said David Linthicum, founder of Linthicum Research LLC and analyst at theCUBE Research. "When you look at what Agentic AI is, it's built on the shoulders of things we've done in the past in building AI agents. The primary difference between Agentic AI and traditional large language model LLM processing like ChatGPT, exists in the operational capabilities and frameworks that they use."
In a CUBE Conversation from SiliconANGLE's studios, Linthicum shared topical insights during the AI Insights and Innovation podcast series. He discussed agentic AI as the future of intelligent systems, offering a more dynamic and adaptable alternative to traditional AI systems.
Contemporary AI systems, especially those underpinned by large language models, have been instrumental in advancing the field. However, these models are limited by their static nature. They generate responses based on their training data and do not actively interact or adapt to external environments. While they can be augmented with external databases and tools, they remain fundamentally constrained by their pre-programmed knowledge.
In contrast, agentic AI is an advanced evolution of LLMs. These AI agents are designed to process information in a more compact and task-specific manner. They integrate tool-calling capabilities, allowing them to gather and utilize up-to-date information, make dynamic decisions and even consult with other agents and databases, according to Linthicum.
"These agents can autonomously perform tasks and adapt to their environment by looking at these external tools or APIs and they're able to collect the necessary information in real-time," he said. "It can be designed to carry out very specific things very well and learn as it goes. In other words, it's an artificial intelligence system, so it can build its knowledge model as it experiences things in terms of processing. Agentic AI systems can also plan and decompose complex tasks into subtasks providing accuracy and efficiency through continuous planning, reasoning and reflection process."
In incorporating human-like reasoning, agentic AI systems require a selection of components; the first of which is perception and sensing. These AI systems gather information from their environments through sensors, cameras and data streams. They can operate autonomously, making decisions based on real-time data inputs. For example, an AI agent in a pump jack can monitor temperature, production rate and maintenance needs, making adjustments as necessary to ensure optimal performance, Linthicum explained.
The second component is information processing. Using algorithms and neural networks, AI agents process and analyze data to make informed decisions. These systems are capable of continuous learning and adaptability, refining their behavior based on feedback and new information.
"These things can make decisions with insight into decision-making frameworks, rule-based systems, machine learning frameworks and reinforcement learning frameworks," Linthicum said. "Again, they can function like an LLM. Most of them are going to be small language models."
The third component is action execution. AI agents can execute actions through robotic actuators or software commands. They can perform tasks autonomously, such as controlling the operations of autonomous vehicles, managing industrial automation processes or assisting in healthcare settings, according to Linthicum.
The final component, and key differentiator from traditional AI, is learning and adaptability. These systems can build their knowledge models over time, improving their performance based on experiences.
"If it's an agent that's running in a camera, it knows that if the humidity is too high, that the camera is going to be cloudy and how to clear that just by being told how to do it one time," Linthicum said. "And then next time the humidity is high and the camera lens is clouded up, it knows how to clear that."
Agentic AI lends itself to many cutting-edge use cases, from autonomous vehicles to healthcare, industrial automation and personal assistance. In healthcare, it can be used for personalized medicine, robotic surgery and patient monitoring systems. AI agents can monitor vital signs, relay information to healthcare providers, and help manage patient care more effectively, according to Linthicum.
"We have those kinds of applications that are emerging today, many of which will be AI agent-based," he said. "Industrial automation, smart manufacturing, predictive maintenance, supply chain optimization -- I use that personally myself. Anything that can benefit from the utilization of an intelligent agent that's already able to carry out a certain narrow set of duties, that's really what it's good at."
Here's theCUBE's complete episode with David Linthicum: