The next frontier in artificial intelligence is being touted as "agentic AI." It can autonomously solve complex, multi-step problems using sophisticated reasoning and iterative planning. This is expected to improve productivity and operations across industries.

NVIDIA announced on the 31st that it is providing various tools and software to help companies build Agentic AI.

Agentic AI systems collect vast amounts of data from multiple sources, independently analyze problems, and develop strategies. They also perform tasks such as supply chain optimization, cybersecurity vulnerability analysis, and assisting physicians with time-consuming tasks.

■ Agentic AI operation process

Agentic AI uses a four-step process to solve problems: recognition, inference, action, and learning.

AI agents collect and process data from a variety of sources, including sensors, databases, and digital interfaces. The recognition process includes tasks such as feature extraction, object recognition, and identification of relevant objects in the environment.

The Large Language Model (LLM) then serves as an inference engine, understanding the task and generating solutions, while also tuning specialized models for specific functions such as content creation, vision processing, and recommendation systems. The inference stage uses techniques like retrieval-augmented generation (RAG) to access proprietary data sources and deliver accurate and relevant output.

The action process allows agentic AI to quickly execute tasks according to established plans by integrating with external tools and software through application programming interfaces (APIs). Furthermore, guardrails can be set to ensure AI agents execute tasks correctly.

For example, a customer service AI agent can only process claims up to a certain dollar amount, and claims exceeding that amount require human approval, thus executing AI tasks within guardrails.

The learning process is continuously improved through a "data flywheel," where agentic AI feeds data generated from feedback loops or interactions into the system to improve the model. This ability to adapt and evolve more effectively over time provides businesses with a powerful tool for better decision-making and operational efficiency.

Strengthening Agentic AI Based on Corporate Data

Generative AI transforms organizations across industries and functions by transforming massive amounts of data into actionable knowledge, empowering employees to work more efficiently.

AI agents access diverse data through an accelerated AI query engine, processing, storing, and retrieving information to enhance generative AI models. A key technology for this is RAG, which enables AI to leverage a wider range of data sources.

Therefore, AI agents learn and improve by feeding back data generated through interactions to the system, creating a data flywheel. This process contributes to refining the model and enhancing its effectiveness.

The end-to-end NVIDIA AI platform, including NVIDIA NeMo microservices, provides the ability to efficiently manage and access data—a key element for building responsive agentic AI applications, NVIDIA explains.

■ Widely used in content, software, and medical fields

The potential applications of agentic AI are vast, limited only by creativity and expertise. From simple tasks like content creation and distribution to more complex use cases like orchestrating enterprise software, AI agents are transforming industries.

AI agents are expected to accelerate their convergence across various industries, including customer service, digital humans, content creation, software engineering, and healthcare.

First, AI agents are improving customer support by enhancing self-service capabilities and automating routine communications. More than half of service professionals reported significant improvements in customer interactions, resulting in shorter response times and higher customer satisfaction.

Interest in digital humans is also growing. These embody a company's brand and provide realistic, real-time interactions, helping sales representatives answer customer inquiries or directly resolve issues during high call volumes.

In the content creation space, agentic AI supports the rapid creation of high-quality, personalized marketing content. Using generative AI agents, marketers save an average of three hours per piece of content, freeing them to focus on strategy and innovation. By streamlining content creation, companies can maintain their competitive edge and improve customer engagement.

In software engineering, repetitive coding tasks are automated to improve developer productivity. NVIDIA predicts that by 2030, AI will automate up to 30% of developers' work time, freeing them to focus on more complex tasks and drive innovation.

Finally, for doctors analyzing vast amounts of medical and patient data, AI agents can extract crucial information and help them make informed treatment decisions. By automating administrative tasks and recording clinical records during patient visits, physicians can reduce the burden of time-consuming tasks and focus on developing relationships with patients.

Meanwhile, NVIDIA is providing sample applications, reference code, sample data, tools, and comprehensive documentation through the NVIDIA NIM Agent Blueprint to accelerate the adoption of generative AI-based applications and agents.

NVIDIA partners, including Accenture, are helping enterprises deploy agentic AI with solutions built with NIM agent blueprints.