Cybersecurity and governance gaps threaten enterprise readiness – Livemint

While 60% of Indian firms report positive AI returns, gaps in cybersecurity, infrastructure and governance threaten wider adoption, shows study.
Indian enterprises are beginning to see tangible returns from artificial intelligence (AI) investments, but weaknesses in cybersecurity, infrastructure and governance are raising questions about their ability to scale adoption sustainably, according to Kyndryl’s third annual Readiness Report published on Tuesday.
The report found that 60% of Indian organisations investing in AI and machine learning reported positive returns, up three percentage points from 2025. However, the proportion completely prepared to manage external risks fell eight percentage points to 27%. The study surveyed 3,700 senior business leaders across 21 countries, including 200 in India.
AI adoption is moving beyond experimentation, with 85% of Indian organisations deploying AI in workflows beyond pilot projects. Yet 70% said they had more AI pilots than they could realistically scale, while 58% reported that innovation frequently stalled after proof of concept.
Infrastructure remains a significant constraint. Although 71% of Indian leaders rated their IT infrastructure as mostly or fully modernised, 59% said significant re-architecture was necessary to support AI at scale. Half of the respondents had substantially changed their AI strategy over the past year, while 31% reported an AI investment failure.
The findings echo a September 2026 ServiceNow Enterprise AI Maturity Index, which found that only 22% of Indian enterprises had AI testing, auditing and risk-assessment processes in place. While 54% were deploying AI agents, just 11% had moved to autonomous workflows. The study also found that 74% of organisations identified data accuracy, access and management as areas requiring improvement.
ForCIOs and CISOs, the implications extend beyond conventional network protection, said Kim Basile is the Chief Information Officer (CIO) at Kyndryl. “As AI agents gain access to enterprise data, applications and business workflows, organisations need to strengthen identity and access management, establish clear permissions for autonomous systems, and ensure continuous monitoring and auditability,” she said.
Cybersecurity was the leading external concern in the Kyndryl study, with 68% of Indian respondents identifying cyberattacks as a major risk. Yet only 35% said they were completely prepared to manage them, a 10-percentage-point decline from 2025.
The October 2026 PwC 2027 Global Digital Trust Insights survey reinforces this concern. Attacks targeting AI systems emerged as the cyber threat organisations were least prepared to address globally. Only 39% had fully formalised operational continuity plans specifically addressing cyber threats.
The challenge of moving AI into production also requires technology leaders to rethink enterprise architecture. In an April 2026 interview with Mint CIO Circle, Venkat Achanta, chief technology, data and analytics officer at TransUnion, highlighted the difficulty of governing, securing and scaling AI agents, particularly in regulated industries.
“But I have the foundation and guardrails, and the agent sitting on my platform uses all of them. That’s what gives us power,” he said, describing TransUnion’s approach to building an enterprise AI platform.
The Kyndryl findings also point to wider organisational challenges, with 37% of Indian respondents reporting disagreements between chief human resources officers and technology leaders over AI strategy.
For CIOs and CTOs, the next phase of AI adoption will therefore require more than additional pilots or infrastructure upgrades. As Basile puts it, it will depend on connecting fragmented systems, strengthening data foundations, embedding security and governance into AI architecture, and aligning technology investments with business and workforce priorities.

source
This is a newsfeed from leading technology publications. No additional editorial review has been performed before posting.