Agentic AI: A new iteration of AI is now taking the world by surprise

As the world continues to explore the potential of generative AI, a new paradigm is emerging in the field of artificial intelligence — Agentic AI. This innovation is poised to redefine how AI interacts with both enterprises and individuals, introducing unprecedented levels of autonomy and decision-making capability.

 

Agentic AI refers to systems capable of acting independently, making decisions with minimal human supervision. These AI agents can perform complex tasks across various domains, from customer service to sales and marketing, opening new possibilities for automation.

 

Industry leaders are taking note

 

At Nvidia’s recent Q3 earnings call, CEO Jensen Huang highlighted the transformative potential of Agentic AI. “AI agents in the future will reason and interact with humans in more meaningful ways,” Huang said, emphasising Nvidia’s commitment to leading this space.

 

Research firms like Gartner also predict rapid adoption. A recent report says, “By 2028, 33 per cent of enterprise software applications will include Agentic AI, up from less than 1 per cent in 2024.”

 

Agentic AI is also gaining traction among Indian startups, which are leveraging it to revolutionise operations for large clients. GupShup, a conversational AI platform, sees immense potential in transforming traditional chatbots into highly capable virtual assistants.

 

“Agentic AI enables a human-like experience on the front end — using text, voice, and multimodal interfaces —while the backend leverages large language models (LLMs) to process and query complex data,” said Gaurav Kachhawa, Chief Product Officer at GupShup.

 

Another pioneer is KOGO, a deep-tech AI firm providing Agentic AI solutions to clients such as Michelin, KTM, and the armed forces of India.

 

“KOGO’s AI agents assist in real-time during customer service calls, offering actionable suggestions, retrieving data proactively, and improving query resolution efficiency. This cuts average handling time by 45 per cent,” said Raj K Gopalakrishnan, Co-Founder and CEO of KOGO.

 

One of KOGO’s key offerings include what it calls “the world’s first AI Agent Store”, offering businesses access to hundreds of AI tools, agents and plugins.

 

Use cases

 

Agentic AI allows companies to automate complex workflows end-to-end across various business functions, be it HR, customer support, or sales and marketing. Unlike traditional GenAI applications, which assist humans only in point tasks, Agentic AI can autonomously own and execute entire workflows.

 

“This has huge implications operationally. It allows companies to realise savings through tasks taking less time, achieving more growth and revenue without adding headcount, and allows staff to work on more high-value, strategic work,” said Phaneesh Gururaj, Head of Engineering-India at enterprise AI startup Ema.

 

Ema’s AI agents, which it calls ‘employees’, can, for instance, act as a recruiter, managing workflows like resume filtering, prioritising and reaching out to candidates autonomously.

 

Tangible returns

 

While it is still early days for Agentic AI, companies using this technology have already begun to report tangible operational improvements. Instead of traditional software tools, which charge for access to software by seat or subscriptions, Agentic AI allows enterprises to “pay for the work actually executed.”

 

“Our early customers using Ema’s Customer Support AI employees have reported 10-15 per cent gains in customer satisfaction scores with more than 80 per cent human agent time saved. Some have already moved more than a third of their customer support team members to other, more business-critical functions,” said Gururaj.

 

Likewise, Gopalakrishnan says that KOGO has seen its enterprise customers report a reduction of up to 50 per cent in operating costs in departments leveraging Agentic AI. Meanwhile, their customer satisfaction scores have gone up by 25-30 per cent.

 

“On average, our enterprise customers have experienced an uplift in overall efficiency of about 30-40 per cent, which has led to improved sales cycles, conversion, cost reduction, and so on… In the short-term, revenues are expected to be impacted by 15-18 per cent,” he said. Kachhawa added that more than 50 per cent of its client base are interested in how they can automate their existing business processes using AI. “When it comes to AI we are presenting Agentic AI as a solution because it is more powerful than classical AI,” he added.

 

Ensuring accuracy

 

The key differentiator between traditional process automation and Agentic AI automation is that AI agents possess cognitive abilities, which entails the ability to make independent decisions. As such, human intervention is further reduced in the process.

 

Therefore, to ensure accuracy, companies like Ema are using “a combination of continuous machine learning models, enterprise-specific custom training, and rigorous data governance protocols.” “Ema’s Fusion Model, a mixture of expert models, intelligently combines over 100 LLM models and many smaller models we built in-house to maximise accuracy at the lowest possible cost for all common tasks in the enterprise,” said Gururaj.

 

Enterprises also employ a “human-in-the-loop” mechanism when it comes to factors such as compliance, brand alignment, continuous training and improvement, or any critical decisions. “While we have eliminated 90 per cent of the human effort, there is always a human in the loop. We currently do not advocate for that level of autonomy,” said Gopalakrishnan.

 

“Moreover, agents are essentially grounded in the data given to them, which belongs to the enterprise. Agents do not use the data sources of any LLM or fine-tuned model. The data is always given by the enterprise,” he added.