At HCLTech’s Noida campus, leaders from industry and technology came together ahead of the India AI Impact Summit to examine a challenge that cuts across geographies and sectors: how do organizations remain resilient when AI itself is accelerating disruption?

Today’s enterprises face simultaneous pressures, including geopolitical volatility, supply-chain fragility, regulatory complexity, sustainability constraints and an unprecedented pace of technological change. Against this backdrop, resilience is no longer just about business continuity or IT recovery. It is about adaptability: how fast organizations can sense change, learn, experiment and respond without breaking trust or economics.

Opening the discussion, Piyush Saxena, SVP and Global Head – Google Business Unit at HCLTech, said: “In the last 12 to 18 months, AI has moved beyond chatbots to Agentic AI, Physical AI and intent-driven development. By 2026, we are entering a year of truth...a year of reckoning...where the focus will be on return on investment and real proof points.”

Joining him were Dipakshi Mahendru, Director, Government Affairs at Intel India, and Vijay Raghavendran, Distinguished Architect/CxO Advisor at Cisco India. Their conversation revealed that resilience in the AI era is not a single capability but a system of choices spanning technology, people and governance.

From automation to autonomy: The capabilities shaping resilience

For Dipakshi, the foundations of AI-driven resilience rest on three converging capability shifts.

“Three traits really stand out. The first is increasing autonomy in systems,” she said. Unlike earlier automation, today’s AI is moving into complex, non-linear domains that require judgment and decision-making, not just execution.

The second shift is advanced reasoning. Rather than replacing humans, AI is increasingly augmenting learning and decision-making, improving the quality of outcomes. This is amplified by multimodality. “AI is no longer limited to a single model or format. We’re seeing seamless integration across text, video, images, voice and audio,” noted Dipakshi, which is an important factor in how organizations absorb information during moments of disruption.

The third capability is where intelligence resides. “AI at the edge is becoming increasingly powerful,” she said, pointing to cost efficiency, energy savings, privacy and lower latency. Edge intelligence distributes resilience, placing decision-making closer to individuals and frontline operations rather than relying entirely on centralized infrastructure.

Crucially, Dipakshi emphasized that these capabilities redefine the human-machine relationship. “Increasingly, we are looking at AI that may surpass human IQ in specific domains,” she said, adding that resilience depends on knowing when and how to engage that partner effectively.

Readiness is layered: Strategy, infrastructure, data and talent

Vijay grounded the discussion in execution realities. “Resilience is about whether an organization can handle the disruption coming from AI,” he said. In Cisco’s experience, many organizations rush to tools before establishing clarity of intent.

He stressed that resilience begins with strategy. Without a clear understanding of what AI is meant to achieve, adoption stalls or misfires. Infrastructure is the next constraint. “Scaling AI isn’t just about GPUs. You need low-latency networks and lossless fabrics,” he noted, highlighting how networking is often underestimated in AI readiness.

Data remains foundational. “Without the right data, it’s a classic garbage-in, garbage-out problem,” said Vijay. Talent is equally critical. “Upskilling is essential, and it has to be supported by the organization,” particularly as AI shifts roles rather than simply eliminating them.

Security and governance must span the full lifecycle. “Not just at runtime, but during development as well,” he emphasized, especially as AI systems move closer to autonomy.

Resilience metrics: Availability, redundancy and human accountability

As organizations move AI into production, the panel agreed that resilience must be measurable. Vijay highlighted dependency risk as a core concern. “If you’re dependent on a public LLM and lose access, will your operations fail?” For mission-critical use cases, he argued, enterprises must plan for availability levels approaching “99.99 availability.”

Governance must come before deployment, not after failure. Responsible AI practices, red-teaming and clear boundaries around enterprise RAG systems are essential to maintaining trust.

He also introduced an important nuance: autonomy is not binary. Borrowing from telecom and automotive analogies, Vijay described graduated levels of autonomy. Most organizations today operate between levels two and three, where AI assists and recommends, but humans remain accountable. “Once you move to higher levels, the agent is doing the full job. And if it fails, there is no neck to choke,” he said, underscoring why human-in-the-loop models remain central to resilience.

From reactive to predictive: The long tail of resilience

One of the most forward-looking insights came from Vijay’s distinction between reactive and predictive AI. “Today, AI is very good at reacting when something breaks,” he said. True resilience, however, comes from identifying weak signals early.

He described focusing on the “long tail,” minor anomalies or low-severity indicators that don’t trigger immediate alarms but signal future failure. By detecting patterns early, AI systems can act proactively, rerouting traffic, adjusting systems or preventing outages altogether.

This shift, from fixing problems to preventing them, represents a fundamental change in how resilience is defined. It also requires organizational patience and leadership support, as long-tail innovation does not always deliver immediate ROI.

Sustainability, cost and supply chains: Resilience beyond technology

Dipakshi expanded the lens beyond technology to economics and sustainability. “AI infrastructure is energy-intensive, which makes total cost of ownership a critical metric,” she said. Resilient models differ by geography; approaches that work in colder regions may not scale in hotter climates like India.

She also tied resilience to global supply chains. Diversification in semiconductor manufacturing is not just a geopolitical issue; it is a resilience imperative. “If there is a disruption, how quickly can you resume business as usual?” she asked.

Culture, trust and psychological safety

Technology alone does not create resilience. Culture determines whether it sticks. Dipakshi reframed trust as psychological safety. “There is a very real fear of people losing jobs,” she said. Addressing that fear requires reframing AI adoption.

Her suggestion was practical: “Redesign work by identifying tasks within a job that can be eliminated, rather than eliminating the job itself.” That shift moves the conversation from threat to productivity.

Leadership capability matters as much as technical skill. “Is senior management trained to help people transition?” she asked. In periods of rapid change, the quality of leadership often determines whether uncertainty becomes paralysis or progress.

Vijay echoed the danger of overpromising. Early deployments without clean data or controls often deliver “around 60% efficacy,” which can permanently damage trust. His advice: start small, deliver 90-95% reliability and scale autonomy gradually.

Resilience as a learned behavior

As the discussion closed, Dipakshi returned to first principles. “Resilience is about getting back up when you’re down,” she said. In an AI-shaped world, resilience comes from lifelong learning, using AI to enhance productivity without losing human judgment and from governance that is embedded into innovation.

Vijay extended the vision outward, pointing to ecosystems of partners, agentic collaboration and cross-domain innovation. The real opportunity, he suggested, lies not just in efficiency but in expanding what organizations can do.

The message from Noida was clear: resilience in the AI era is not driven by autonomy alone. It emerges when organizations align autonomy with governance, economics and human trust. As Piyush noted, the next phase of AI will be judged by “return on investment and real proof points.” Those who treat resilience as a core capability, not a contingency plan, will be best positioned to adapt, compete and lead.