Agentic AI: Promises and perils of letting the machines run the show

As platforms like ChatGPT, Grok, and Gemini have roped-in international attention to generative AI, a new and more autonomous form of artificial intelligence (AI) is making its way into mainstream existence — Agentic AI. This new branch of AI will essentially give machine learning models to take human-like decisions in real time to finalise decisions and solve problems.

Generative AI models are limited to the data they’re trained on. In contrast, agentic AI is capable of learning engagement and decision-making patterns, allowing it to perform tasks without human supervision, a blog by IBM explains. The New York-based tech giant noted in its blog that agentic AI offers more advantages compared to generative AI.

Pros and cons of agentic AI

Once deployed, agentic AI systems are able to perform minor day-to-day tasks as well as take decisions and execute large-scale operations.

For instance, if an agentic AI is asked to place an order for eggs — it will enter a multi-layer approach to execute the order. It will create a plan, find marketplaces that sell eggs around the location, give a delivery address, and process a payment. At each step, the AI system will update the user who initiated the action.

Agentic AI, once coming into a widespread adoption, will help enterprises achieve their network expansion goals — eliminating the need for companies to maintain large teams of human employees. If, for example, one agentic AI agent can make 1,000 decisions a day for a company — deploying 10,000 AI agents across a logistics network could generate 10 million decisions everyday — without requiring any human oversight.

The use of agentic AI, as per PwC, is expected to bring more efficiency to the management of complex, multi-layer workflow systems with intelligent insights and reduced time requirements — at a lesser cost.

While this agentic AI technology brings multiple promises to the table, it also carries numerous perils, blockchain analyst and author Harshajit Sarmah in conversation with CoinHeadlines.

“If agents have financial control and no kill switch, they can’t be recalled or governed,” he said.

Sana Afreen from Abyro Capital, who is also building a decentralized AI hub called PexuAI, also seconded Sarmah while also asseting that provenance over agentic AI is crucial.

“We’re looking at autonomous agents managing inventory, executing payments, negotiating contracts — not just talking about chatbots booking appointments,” said Afreen. “As agentic AI systems start making decisions and transacting on our behalf, its crucial to know who created the agent, who trained it and who would be accountable when it acts.”

Blockchain to the rescue

In order to keep track of the activities performed by agentic AI, blockchain integration will become key, both Afreen and Sarmah told CoinHeadlines.

Blockchain can log an immutable trails on the tasks of agentic AI platforms while also logging details on where an action originated from. As Afreen explained, without a blockchain base, agentic AI becomes a black box with a credit card.

Dr. Preet Deep Singh, faculty of blockchain at the prestegeus IIT-Kanpur, told CoinHeadlines that agentic AI will require blockchain to assign liability to the tasks facilitated by agentic AI tools.

“Knowing what all happened to correct mistakes without any data manipulations would be where blockchain would come in for agentic AI,” Singh noted. “WIth blockchain there would be an immutable audit trail for all the steps an AI agent took. This security would allow for widespread adoption.”

The experts also collectively highlighted that while blockchain integration will improve the coordination and audit ability of agentic AI tools, full decentralization, especially with autonomous agents, is a serious risk. This means, developers of this autonomous technology would need to prioritize on introducing elements of trust, traceability, and control to encourage more adoption.

As per data by Markets and Markets, the agentic AI market is estimated to grow from $7.06 billion in 2025 to $93.20 billion by 2032. Companies like OpenAI, Google, Nvidia, and Microsoft are reportedly exploring the agentic AI sector.