The development of artificial intelligence is shifting from using copilots and vibe coding to creating AI systems capable of reasoning within workflows, coordinating specialized agents, and achieving results with minimal human involvement.
At the DevSparks Hyderabad 2026 event, Elango Somasundaram, Managing Director and CTO of UBS India, noted that the transition to agentic AI is less about replacing engineers and more about changing the very nature of engineering work.
According to Elango, the next competitive advantage will depend not only on improving the models themselves but also on richer context, robust guardrails, and developers who can think beyond just code.
He emphasized: 'We are entering an era of agentic engineering, where AI agents do not just respond to prompts, but understand the context of the task and use it to make more informed decisions.'
Elango traced the evolution of AI through several stages: from robotic process automation and cloud technologies to copilots, vibe coding, and specialized agents. While copilots can predict the next line of code, vibe coding allows programs to be created through natural language prompts, and agents perform specific tasks. Agentic AI goes further: it starts with an intent, determines the necessary tasks, delegates them to specialized agents, and provides a final result.
Agentic engineering is named as the next frontier. Instead of developers manually breaking down requirements into epics, defining dependencies, coordinating teams, and connecting workflows, intelligent systems will increasingly understand these interconnections autonomously. This points to a different approach to software development, where AI begins to participate in architectural thinking traditionally performed by experienced engineers.
According to Elango, skills are rapidly becoming commoditized. Most organizations now have access to powerful foundational models and specialized competencies that have been democratized through open-source ecosystems and AI/ML communities. He argued that the distinguishing feature is precisely the context.
'How do we bring more context into the agents we develop? You can get the skill, but how do you ensure the inclusion of context? Data is becoming crucial,' he stated.
For enterprises, especially in regulated industries like banking, context includes proprietary business processes, product knowledge, security policies, customer expectations, and organizational workflows. General models can write code, but they do not automatically understand how a specific bank approves transactions, how a medical facility manages compliance, or how an automotive company designs its products.
This makes data quality fundamental. Enterprises may possess vast amounts of information, yet it is often fragmented across systems, duplicated, inconsistent, or disconnected from the workflows required by AI. Consequently, the focus shifts from the number of employees using AI tools or the volume of tokens consumed, to the ability of AI systems to leverage corporate context to achieve accurate and reliable outcomes.
This also changes the nature of work for developers. Instead of just implementing stories and epics, they are increasingly required to enrich AI systems with the knowledge necessary for correct problem-solving.
Elango explained that agentic AI cannot succeed by focusing only on the application level. He described enterprise architecture as a set of interconnected layers covering infrastructure and network interaction, identity management, software development practices, governance, and security. A modern application can run in both cloud and on-premise environments, connect to dozens of services, and function within organization-specific architectural principles. Therefore, agentic systems must reason within the entire broader ecosystem, not just generate functional code.
Identity and access management gains particular importance as agents increasingly exchange information with each other. If inaccurate or hallucinated information is passed from one agent to another, the error can compound and lead to an incorrect and unintended workflow outcome, which can have serious consequences. Therefore, enterprises need guardrails at every architectural layer: connectivity, authentication, software development lifecycle elements, regulatory compliance, and security policies become part of the agentic workflow, rather than end-of-process checks.
Elango also touched upon the changing role of humans. As AI automates more execution stages, developers will continue to define goals, verify results, and provide business context. For enterprises, more sophisticated reasoning models have made AI adoption simultaneously more promising and more complex. He noted: 'While we gain greater efficiency in terms of accelerating code development and even testing, the final result requires rigorous testing. AI can generate tens or thousands of lines of code in just a minute; we must ensure there is nothing wrong with it.'
Previously, AI workflows heavily relied on human code review before moving to testing. Agentic systems add another layer of complexity, as multiple agents can independently generate, validate, and execute different parts of the workflow. As a result, enterprises are slowing down deployment—not because the technology promises nothing good, but because governance has become more important than speed.
Every model entering the enterprise environment now requires careful curation. Security teams must ensure that confidential information does not leak to external models, and developers must be certain that AI-generated outputs comply with internal policies before going into production.
Elango views the banking sector as an early example of how agentic AI can transform operational models. Routine processes such as payment calculations, customer operations, and workflow orchestration can eventually be managed by coordinated agents instead of large operational teams. However, this future still depends on the architects, engineers, business leaders, and operational experts designing the ecosystem behind these agents.
A larger opportunity, in his view, lies in enterprise differentiation. Each industry has a unique context that competitors cannot easily replicate. In banking, this includes regulatory knowledge, risk frameworks, payment infrastructure, and customer trust. As agentic systems mature, these contextual advantages may become more valuable than access to general AI models.
Elango also discussed what agentic AI means for developers. He insisted that experience still matters. An AI model can generate a thousand lines of working code, but an experienced engineer often knows how to solve the same problem more efficiently and effectively because they understand the business context, lessons from partial releases, know about undocumented customer needs, and learn through continuous interaction and refinement.
The next generation of developers, he said, should possess three capabilities: understanding the business problem, collaborating within complex corporate ecosystems, and clearly articulating intentions for AI systems, rather than just writing code. As agentic engineering evolves, coding will become just one part of a developer's toolkit. Context, judgment, and systemic thinking will become increasingly vital for those aiming to lead with AI.