Beyond chatbots, Agentic AI is expected to change significantly how businesses operate. It is especially useful in customer service and other industries, as it is designed to plan and take action on its own using reasoning.
A study by market intelligence firm IDC found that consumers in Asia Pacific are projected to spend $32 billion on AI-driven programmatic shopping by 2028.
“Agentic AI transforms GenAI into actionable services by seamlessly integrating into workflows to improve decision-making and task execution with a high level of independence,” said Deepika Giri, associate vice president and head of AI Research, IDC Asia/Pacific, in an email interview with Back End News. “It also simplifies the underlying model layer, allowing flexibility in model choices without affecting the application layer.”
Key concepts of Agentic AI
Agentic AI builds on AI assistants by using advanced reasoning and step-by-step planning to solve complex problems on its own, according to Nvidia. It aims to boost productivity and streamline operations across industries.
Simply put, Agentic AI systems process vast amounts of data from different sources. They analyze user requests, create strategies, and then take action. In contrast, chatbots only answer predefined questions, which limits their usefulness when customer issues go beyond their training. Agentic systems use various large language models (LLMs) to execute tasks and analyze data.
Nvidia has introduced a set of Agentic AI models that claim to offer “the highest accuracy on a wide range of tasks, outstanding computing efficiency, and an open license for enterprise use.”
Giri explained that, for now, Agentic AI capabilities are being added to SaaS applications to enhance them with AI and GenAI features. In the long run, these agents could evolve into a central intelligence system that connects multiple SaaS applications—if issues related to data security, privacy, and reliability are resolved.
Real-world applications
AI agents are expected to drive automation across industries, including government services. A recent Salesforce report found that 87% of U.S. respondents would use an AI agent to navigate public sector processes.
AWS sees potential in healthcare, where AI agents can process vast amounts of medical data and run on specialized LLMs. Clinical AI agents could analyze doctors’ notes using natural language processing (NLP) to detect critical information, distinguishing between current and past issues. As molecular test data agents, they could decode genomic data from biopsy samples to identify biomarkers for personalized treatments.
Challenges
“It is crucial to design Agentic AI systems with transparency and explainability in mind,” Giri emphasized. “Additionally, human oversight is needed to prevent systemic errors that could have serious consequences.”
Despite optimism about improving efficiency, concerns about job losses, privacy, and governance remain.
If AI agents can make decisions, what happens to industries like call centers? BPOs employ millions of people to handle technical support. If AI agents can answer questions and solve problems, fewer human agents may be needed.
That said, human oversight will always be essential to prevent errors. While some jobs may be automated, new roles will also emerge.
The UC Berkeley Sutardja Center for Entrepreneurship and Technology highlighted that LLMs can produce inaccurate results, which poses risks in critical fields like healthcare.
The center also pointed out that AI systems are vulnerable to adversarial attacks, which exploit weaknesses in their reasoning processes. These attacks can affect different stages of decision-making, from input interpretation to final outputs, potentially leading to flawed results.
The future of Agentic AI
With IDC predicting AI spending to surpass $30 billion by 2027 for personalized customer experiences, it’s clear that Agentic AI is gaining traction.
Earlier this month, AWS announced a new division dedicated to Agentic AI development, led by executive Swami Sivasubramanian. Reuters quoted Matt Garman, CEO of AWS, as saying in an email: “Agentic AI has the potential to be the next multi-billion-dollar business for AWS.”
Nvidia describes it as the next step in generative AI, envisioning enterprise AI agents as the core of AI-driven factories that generate tokens to unlock new levels of intelligence and productivity.
Though still a developing concept, Agentic AI is on track to become as essential to business operations as GenAI.