As Indian enterprises move beyond AI pilots to full-scale deployment, the conversation is shifting from which models to use to how well those models understand the business around them. In this interview, Atul Ahuja, Area Vice President and General Manager, Elastic India, explains why India is outpacing its global peers in enterprise AI adoption and why financial services, under intense regulatory and competitive pressure, are setting the standard for the rest of the market. Ahuja unpacks the rise of context engineering as the new differentiator over raw model capability, the real reasons AI investments fail to translate into measurable outcomes, and why open source alone isn’t enough without operational guardrails. He also outlines what CXOs must prioritise to move AI from experimentation to lasting business impact.
Area Vice President and General Manager
Elastic India
CIO&Leader: What is driving India to emerge as a leader in enterprise AI adoption?
Atul Ahuja: India is rapidly establishing itself as a leader in enterprise AI adoption, with organisations quickly moving from experimentation to practical implementation. Deloitte’s 2026 State of AI report reveals that Indian companies surpass their global peers in AI utilisation, with 40% using AI extensively or across the entire enterprise, versus 28% worldwide. Furthermore, 94% of Indian firms intend to increase their AI investments, indicating a shift from trial phases to the integration of AI into essential business processes.
One of the main drivers of this trend is the evolving mindset around AI. Organisations are moving beyond viewing AI as an experimental endeavour. The conversation has moved to specific problems, productivity, analytics, getting people to the knowledge they need, and whether the investment is paying back. When the question becomes ROI rather than “can we build it,” you know a market has matured.
The financial services sector is leading the charge. Banks, insurers, and capital markets firms are under immense pressure from regulators, competitors, and customer expectations for prompt, reliable service. This urgency demands that AI be robust and enterprise-ready from the start. When the sector that can least afford a wrong answer starts deploying AI at scale, it raises the bar for everyone. The rigour it demands, including grounded outputs, auditability, and tight access control, becomes the standard the rest of the market builds towards.
Underneath both lies the foundation: a supportive regulatory environment and talent. India has the world’s second-largest developer population, and these are people already fluent in the data and cloud environments needed to scale AI effectively. That’s also why India’s Global Capability Centres (GCCs) have become AI-led hubs, with global firms routing real product and engineering strategy through them, giving the country direct influence over how enterprise AI is built, not just where it runs. Put high user acceptance, a bias towards practical use, and that much headroom together, and it’s clear India is early in this, not late.
CIO&Leader: How is context retrieval reshaping the way enterprises extract value from data at scale?
Atul Ahuja: The enterprise data problem is never just about volume and size. The information you need is scattered across applications, repositories, and formats, and finding the relevant piece at the right moment is genuinely hard. Context retrieval closes that gap: it connects AI systems to the most relevant enterprise knowledge in real time, so the output is accurate and usable. As companies scale AI, the goal shifts from “retrieve information” to “deliver the right context for this user, this task, this workflow.”
That’s what’s given rise to context engineering, and I’d argue it’s the capability that now separates enterprises scaling AI reliably from those still stuck in pilots. The practice itself is straightforward to describe: select, organise, and deliver the enterprise data an AI application actually needs, rather than flooding a model with everything and hoping. Do it well, and you get grounded in trusted, business-specific knowledge, which gives you relevance, accuracy, and explainability. A capable model on its own doesn’t get you there. A well-grounded one does.
And the payoff goes well beyond better search. When responses are grounded in a company’s own data and operations, people stop spending hours searching for information, customers and employees get answers that reflect real intent rather than generic guesses, and, most importantly, AI agents finally have the context to carry out complex processes reliably and at scale.
That last point is where this is heading. As enterprises move from experimenting to deploying AI across the business, context becomes the foundational layer of the stack. Success won’t come from access to a larger model. It’ll come from grounding your model in trusted, real-time enterprise data. Context is what turns AI into accurate decisions, reliable automation, and measurable outcomes.
CIO&Leader: Where are organisations struggling to convert AI investments into measurable business outcomes?
Atul Ahuja: The problem isn’t how much organisations are investing; it’s that the investment and the value keep landing in different places. Teams move fast to experiment with models, but when a model lacks the right context, the output is either too generic for real-world use or starts inventing things. Those hallucinations, where the model produces confident but fabricated answers, are exactly what enterprises can’t afford.
In a consumer app, the occasional wrong answer is a minor annoyance. In a bank, it isn’t. If a leader is going to trust AI to inform a real decision, the output has to be dependable, and that only happens when it’s grounded in the organisation’s own and relevant data rather than generic knowledge.
The other pattern I often see is effort directed at the wrong target. A team will generate thousands of help articles and count it as an AI win, but if customer tickets aren’t getting resolved any faster, there’s no real return there. The impact comes from improving the operations that actually matter to the business. And that comes back to the data foundation: you need one that preserves relationships across your structured and unstructured sources, so agents can retrieve the right context and work reliably in production. Get that right, and the accuracy, the safety, and the ROI follow.
CIO&Leader: What role does open-source play in accelerating enterprise AI innovation and reducing vendor lock-in?
Atul Ahuja: Open source gives enterprises the fastest way to start AI initiatives. Teams can spin up environments quickly, work across models, and test use cases without locking into a fixed stack. That flexibility helps them identify where AI delivers value.
The real question is not open source versus commercial, but how to carry that openness into production. As data volumes grow and workloads become business-critical, teams need consistent behaviour, sustained performance, and enforceable security and compliance controls. Openness has to be matched with the operational guardrails production demands.
This is why most enterprises land on a hybrid approach: an open foundation for flexibility, with the control layers required to run reliably at scale. It is the model we built Elastic around. We give organisations an open foundation, then add the capabilities to unify their data, enforce governance, and operate with confidence, so they can move from experimentation to production without trading away control of their stack.
CIO&Leader: What key enterprise technology shifts do you expect to define the India market through 2026?
Atul Ahuja: The next phase in India is less about adopting AI and more about operationalising it. Most enterprises I talk to have already proven the technology works. The harder question now is running it at scale in production, against real business outcomes, and that shift from pilots to production is the story of the next 12 to 18 months. India is moving on it faster than most markets, and the push is coming hardest from regulated sectors like financial services, where the competitive and regulatory stakes leave little room to stay in pilot mode.
Three shifts stand out.
First, agentic AI moves from conversation to action. Enterprises are past being impressed by chatbots. What they want now are systems that can carry out multi-step processes, coordinate tasks across applications, and make decisions with a human in the loop rather than a human doing all the work. Workflow automation is where the return actually shows up.
Second, context becomes the differentiator, not the model. As agents take on real work, the constraint shifts from how capable the model is to how well it understands the business around it. That is why teams are moving past basic RAG toward context engineering: selecting and delivering the exact enterprise data an agent needs, grounded and in real time. The organisations that get this right will pull ahead, regardless of which model they run.
Third, the platform consolidates. Enterprises are tired of stitching together separate tools for search, observability, and security, each holding its own copy of the data. The advantage now lies with those who can bring that data onto a single foundation and see across applications, infrastructure, and security. That is what makes an organisation both resilient and genuinely AI-ready.
The edge over the next 12 to 18 months won’t come from owning more models. It will come from grounding AI in trusted data and wiring it into the processes that run the business, securely and measurably. That is a very different discipline from experimentation, and it is the one Indian enterprises are now building.
CIO&Leader: What should CXOs prioritise today to move AI from experimentation to sustained business impact?
Atul Ahuja: CXOs need to understand that although pilots offer visibility, they don’t automatically alter business operations. To achieve production within months, the focus should be on integration, reliable data, and efficient context retrieval.
I recommend CXOs focus on three pillars:
1. Defining the problem and business workflows needed: Work closely with IT and CTOs to select tools based on specific business outcomes rather than technical novelty
2. Build a unified sovereign data and context layer: AI is only as effective as the data and context it can access. Organisations should focus on unifying structured and unstructured data, enabling accurate retrieval of relevant information and business context across the enterprise to support decision-making and automation.
3. Security and Access: CXOs must ensure that AI tools retrieve data only within strict access controls to maintain security, compliance, and privacy.
Using managed tools, such as the Elastic Agent Builder, can help teams quickly bridge this gap by connecting data to AI components, including vector databases and LLMs, through a secure, natural-language interface. Built natively on the underlying data platform, it provides built-in retrieval, ranking, and access controls without the need to assemble separate pipelines.