By Ravindra Ramnani, Senior Manager, Solutions Architecture, Elastic
In 2026, organisations won’t just deploy AI – they will examine how they can provide AI with as much context as possible, so that it can deliver relevant, accurate, and responsible output. Context engineering, which provides an AI agent with the right data at the right time, will become a critical element for AI success.
India leads the Asia Pacific and Japan region in AI adoption with a 92% adoption rate of AI among employees, according to a recent Boston Consulting Group study. As technology matures, India will need to focus on tangible business results driven by AI, rather than just on adoption rates.
The large language model (LLM) is no longer sufficient on its own. By connecting LLMs with contextually accurate data, AI can review the data, plan, and act to achieve a business outcome with minimum human oversight. Therein lies the challenge: AI needs context to succeed at the tasks it is assigned, and organisations will need to find ways to provide that context.
Trend 1: Decisions backed by a wealth of context
AI agents are designed to independently gather, interpret, and act on information — moving toward a defined goal. Because these agents operate with a high degree of autonomy, their success depends heavily on context engineering.
Context engineering focuses on equipping an AI agent with the most relevant, high-quality information at the exact moment it is needed. That’s not easy. After all, that data is often scattered across countless sources, both structured and unstructured: from databases, documents, and emails to CRM systems, Slack messages, social media posts, log files, and transaction data.
Trend 2: Building agents, block by block
Development of such AI agents doesn’t have to be done on an in-house platform, which costs dearly and takes years to build.
Leading tech companies are now offering pre-built components of AI agents for organisations to pick and choose, similar to “Lego blocks”. This modular approach allows business users and developers to combine functions for a use case, ranging from data retrieval, calculation, and reasoning to guardrails, to assemble complex AI solutions independently.
Suppose a financial firm needs a tool to help private bankers identify clients at risk during a market crash. Instead of writing custom code for months, a developer turns to an Agent Builder. With it, they can assemble the solution piece by piece: a retrieval component to search client portfolios, a calculation engine to score risk, a memory component that knows which clients overlooked previous warnings, and a reasoning engine to analyse the relationship between liquidity and historical data. Finally, they add a guardrails component to ensure the AI never recommends irreversible actions without human approval.
In a fraction of the usual time, the developer deploys a secure, fully functional “Risk Assistant”. The private banker no longer has to sift through complex data manually. They can simply chat with this tool and ask, “Which clients are most exposed right now?” The agent handles the complexity in the background, delivering immediate insights that drive business value.
Trend 3: Speed in the Battle Against AI-Driven Cyberattacks
More powerful AI and the increasing need for rapid access to valuable data inevitably also mean greater security risks. The Elastic 2025 Global Threat Report notes that the deployment of Large Language Models over the past year led to a marked growth in the number of cyberattacks. The speed of the attack is increasingly often the decisive factor.
In short, thanks to their use of AI technology, cybercriminals need increasingly less time to achieve their goal, and the good actors need to move at the speed of AI. Despite advancing automation, many cybersecurity systems still rely on human analysts. By the time they determine exactly what’s happening during a new type of attack, the cybercriminals have often already succeeded.
Successful defense needs the speed of an AI-assisted response. Unlike traditional rule-based systems, adaptive AI-driven software learns incredibly quickly from the behaviour of a new threat. An AI-driven tool like Attack Discovery can therefore predict the objectives and future actions of malicious software, and respond without human intervention to neutralise the threat.
Here too, there’s potentially a major future role for AI agents. From a security analyst building an incident response agent to a data team having its agents search 24/7 for data quality issues.
The New Evolutionary Phase of AI
India’s next big step is the proliferation of AI agents that drive tangible outcomes. Professionals are no longer dependent on standard models or isolated experiments; they now take control by designing, training, and optimising their own Agents. AI transforms from a support tool into a strategic force that accelerates operations, decision-making, and innovation.
Organisations that take this step now will build an advantage that is difficult to match. They will make faster decisions, anticipate risks more effectively, and extract greater value from data. Agentic AI and context engineering form the backbone of the next AI era — a period in which AI is no longer a promise, but a direct driver of measurable business impact.