India’s enterprise AI push is moving beyond experimentation, but the next phase of adoption will depend less on the availability of AI models and more on how organisations manage the data, context, security and costs surrounding them.
That is the view emerging from Elastic’s India leadership, as enterprises increasingly move AI initiatives from pilots towards production-scale deployments.
Atul Ahuja, Area Vice President and General Manager, India, Elastic, said Indian enterprises are showing strong momentum in AI adoption, with several organisations already moving beyond conventional proof-of-concept exercises towards initiatives tied to specific business outcomes.
“The answer lies in how you deploy the architecture and how you put the right data to use,” Ahuja said.
For enterprises, however, that is easier said than done. Organisations today hold data across multiple formats and environments — from structured transactional databases and PDFs to scanned documents, images, videos, audio and legacy records. Bringing this information into an AI workflow while preserving its context has emerged as one of the biggest challenges.
Ahuja believes the quality and relevance of the underlying data will ultimately determine whether an AI initiative delivers meaningful business value.
“Every organisation today has exploded with the volume of data,” he said, adding that enterprises need to be able to bring the right data and context together to generate useful outcomes from AI.
From data overload to contextual AI
The challenge is no longer simply storing enterprise data. It is being able to find and interpret the right information quickly enough for AI applications to use it effectively.
Ahuja cited examples ranging from scanned documents and expense statements to transactional records, audio and video. The definition of enterprise data, he argued, is expanding rapidly, making the ability to search and retrieve relevant information across these formats increasingly important.
This is where Elastic sees search and retrieval becoming a critical layer in enterprise AI architectures.
Rather than forcing organisations to consolidate all their data into a single repository, Elastic’s approach allows enterprises to search information across existing sources using connectors and APIs.
“You don’t have to put all your data actually into one place. Elastic can search it wherever it is,” Ahuja said.
For CIOs, the proposition is particularly relevant as organisations look to accelerate AI deployment without undertaking lengthy data-migration or consolidation projects.
The objective is to make enterprise data accessible to AI systems while retaining the existing architecture and sources where practical.
The token-cost problem
As enterprises experiment with AI agents and increasingly autonomous workflows, another concern is emerging: the cost of inference.
The proliferation of AI agents means enterprises could potentially consume large volumes of tokens through repeated queries, lookups and interactions with language models. For organisations running AI at scale, that could quickly become a significant operational expense.
Ahuja said architecture and context engineering will play a crucial role in controlling this cost.
“If I am prompting something, it doesn’t go into an endless loop and try and use so many more tokens before it can find the right answer,” he said.
The ability to retrieve relevant enterprise information efficiently can reduce unnecessary interactions with language models, potentially improving both productivity and cost efficiency.
This is also where the quality of enterprise data becomes critical. Poorly organised or irrelevant information can result in inefficient retrieval, inaccurate responses and higher token consumption.
“It’s garbage in, garbage out,” Ahuja said.
For CIOs, therefore, the AI conversation is increasingly shifting from which model to use towards how the broader architecture can ensure that models receive the right information, in the right context, at the right time.
Security remains a non-negotiable guardrail
The other major consideration is security.
As enterprises connect AI systems to increasingly sensitive internal information, access controls and governance become critical. Financial institutions, government organisations and other highly regulated sectors cannot afford to allow sensitive information to flow beyond authorised boundaries.
Ahuja said Elastic’s security portfolio has become one of its fastest-growing areas, with the company positioning its capabilities around security information and event management, extended detection and response, and an agentic approach to security operations.
The broader objective is to use AI to reduce the time required to detect, investigate and respond to security incidents across large volumes of events, logs and telemetry.
For enterprises, the attraction is not simply automation. It is the ability to process massive volumes of security information while shortening response and resolution times.
Sovereignty is another important consideration, particularly in India.
Ahuja said enterprises increasingly want control over where their data and security intelligence reside. Elastic therefore supports both cloud and customer-controlled environments, allowing organisations to deploy the technology within their own infrastructure where required.
BFSI remains a major AI and observability market
Elastic’s focus on India also reflects the growing technology requirements of the BFSI sector.
Ahuja said the company’s footprint extends across banks and insurance companies, with use cases spanning underwriting, claims management, customer service, observability and regulatory compliance.
In banking, observability has become particularly important as digital services become more complex.
A failed or slow transaction can originate from multiple layers — from the network and infrastructure to an application or a specific piece of code. The ability to trace telemetry across these layers can help organisations identify the source of a problem more quickly.
For banks operating highly regulated environments, traceability also supports audit and compliance requirements.
Beyond BFSI, Elastic is seeing demand from digital-native companies, large enterprises, healthcare organisations and the public sector.
Ahuja said the company has also worked with government and defence organisations, where data sovereignty and security requirements make the ability to operate within controlled environments particularly important.
The enterprise AI stack is changing
The larger message from Elastic’s India leadership is that AI adoption is moving into a more demanding phase.
The first phase was about experimentation — testing models, building pilots and understanding what generative AI could do.
The next phase is about engineering those capabilities into production systems.
That requires enterprises to solve several interconnected problems: making fragmented data accessible, preserving context, controlling inference costs, enforcing security and governance, and ensuring that AI systems can operate reliably at scale.
For CIOs, the AI question is therefore becoming broader than selecting a foundation model.
The competitive advantage could increasingly come from the architecture underneath it — and from an organisation’s ability to make its own data searchable, contextual and usable by AI.
As Ahuja put it, the combination of the right data and the right engine can help organisations get “the right information” while keeping costs and guardrails under control.
For Indian enterprises racing towards agentic AI and production-scale deployments, that foundation may ultimately determine whether today’s AI initiatives become sustainable business systems or remain expensive experiments.