Trust Issues Persist as Poor Data May Fail Agentic AI
Trust and accountability were identified as critical concerns, particularly in regulated sectors such as healthcare and finance, where organisations must be able to explain decisions and establish responsibility when errors occur.
Hyderabad: Agentic AI is being widely used in banking, healthcare, manufacturing and software development, but experts claim concerns around trust, governance, accountability and data readiness continue to shape its adoption.
Speaking at an industry-academia roundtable organised to examine the preparedness of organisations to deploy Agentic AI, experts said Agentic AI systems can plan, reason, use tools, remember context and perform tasks with a degree of autonomy, unlike conventional chatbots that respond to prompts. However, participants stressed that enterprises are not looking to hand over complete control to AI systems.
“Think of it as an autonomy slider rather than an on-off switch,” said Prof. Karthik Vaidhyanathan. “An agent can draft an email, recommend a supplier or analyse a loan application, but organisations still want humans making critical decisions.”
The roundtable discussed emerging applications including fraud detection and customer outreach in banking, faster healthcare authorisation processes, insurance document management, inventory and procurement analysis in manufacturing, and support for software development tasks such as testing, verification and documentation.
Participants noted that the biggest hurdles are often not the AI models themselves but the systems surrounding them. Unlike traditional software, AI agents may complete tasks while relying on incorrect information, outdated data or unsuitable tools, making evaluation and oversight more complex.
Trust and accountability were identified as critical concerns, particularly in regulated sectors such as healthcare and finance, where organisations must be able to explain decisions and establish responsibility when errors occur.
Experts also pointed to poor-quality data as a major obstacle to successful deployment. Many organisations, they said, discover that strengthening data systems becomes a prerequisite for effective AI adoption.
The conference examined the rise of “vibe coding”, where developers increasingly describe requirements while AI tools generate code. While such tools can improve productivity, participants stressed that technical expertise and human judgment remain essential.
Summing up the discussions, Prof. Vaidhyanathan said organisations must balance the opportunities offered by Agentic AI with risks related to governance, cost and sustainability, rather than treating the technology as either a cure-all or a passing trend.
