India’s agentic AI dream for fintech needs an infrastructure relook | Artificial Intelligence News

Agentic commerce is expected to evolve in three phases: assistive, guided, and eventually autonomous flows.
Current infrastructure, built for human interactions and deterministic software, is not adequate for autonomous AI agents, highlighting the need for a new layer of infrastructure for permissions, authentication, authorization, auditability, and liability.
To support AI agents in shopping, transacting, or negotiating on behalf of users, there must be clear consent, spending controls, regulatory clarity, and discoverability that is equally usable by both humans and machines.
The deployment of multiple AI agents for various tasks raises significant challenges around edge cases, accountability, and how existing payments and commerce infrastructure will handle non-human interactions.
Banks and financial systems need to build an intermediary software layer to validate agents’ access, ensure data privacy, and enforce security policies, creating audit trails before accessing information.
New forms of fraud, like deepfakes and tampered documents are already present and evolving, requiring risk-based authentication to mitigate these threats.
The industry continues to focus on rethinking and strengthening the infrastructure that supports agentic AI, while addressing security, control, and permission concerns as adoption moves beyond initial pilots.

You may have missed