Enterprises seeking to transition from generative AI (GenAI) pilots to agentic AI systems must reassess processes, governance, observability, and accountability, according to technology leaders at the ETCIO Annual Conclave 2026. As AI progresses from user assistance to autonomous decision-making, the speakers emphasized the importance of balancing speed, trust, and operational control to achieve measurable business outcomes.
This discussion took place during a session titled “From GenAI to Agentic AI: What It Really Takes to Move from PoC to Business Impact,” moderated by Shantheri Mallaya, Editor, ETCIO. Mukul Jain, CTO of Axis Max Life Insurance, noted that organizations are increasingly shifting from AI-assisted workflows to systems capable of independent action and orchestration. However, he highlighted that customer-facing and regulated sectors like insurance necessitate a gradual evolution with robust governance and human oversight.
Jain provided an example involving a multi-agent system managing customer policy queries, wherein one AI model drafts responses while another validates them before they are communicated to customers. He stated, “Human-in-the-loop is not a weakness; it is an operating model during this transition journey.”
He further stressed the need for enterprises to establish clear boundaries surrounding the operational independence of autonomous systems versus areas requiring human review.
Viral Davda, CIO of BSE, emphasized the necessity for organizations to clearly differentiate between workflow automation and genuine agentic AI. He recommended that AI implementations should begin with measurable Key Performance Indicators (KPIs) and defined business outcomes before scaling operations further. He referenced BSE’s AI-driven listing compliance platform, which has expedited processing timelines for listing applications from 30-45 days to just one to three days. “If there is decision-making involved and measurable outcomes attached to it, then you are entering the world of agentic systems,” Davda remarked.
Davda also indicated that enterprises must overhaul their governance frameworks for autonomous AI, as existing controls designed for traditional software systems are inadequate.
Himanshu Pant, CDO of Adani Group, warned that organizations cannot effectively scale agentic AI atop flawed workflows or fragmented data systems. He advocated for resolving foundational processes and establishing robust data infrastructures prior to integrating layers of autonomous decision-making. “If the processes are not right, AI will only accelerate the error,” Pant stated.
Accountability and trust emerge as critical issues as AI agents edge closer to managerial and decision-making roles within companies. Bharani Subramaniam, CTO of Thoughtworks for India and the Middle East, noted that many enterprises still describe orchestrated workflows as agentic AI, despite remaining in the realm of deterministic automation. True agentic AI, he explained, materializes when systems tackle complex research, optimization, or uncertain problems where predefined steps do not exist.
He emphasized the necessity for observability, reversibility, and machine governance, urging enterprises to engineer systems capable of detecting, auditing, and rectifying incorrect AI decisions when failures occur.
The panel collectively asserted that AI adoption should not be perceived as isolated experimentation. Instead, advancing agentic AI requires process redesign, concrete KPIs, enhanced tooling, improved data quality, governance frameworks, and operational resilience.
The session concluded with a consensus that the future of enterprise AI should center around autonomous systems capable of responsible decision-making. Leaders reiterated that the successful scaling of AI hinges on embedding trust, observability, and governance directly into organizational operating models.
(With contributions from Sachi Srivastava).
Published On May 25, 2026, at 09:43 AM IST.
