New Delhi, June 19 -- As enterprises move from experimenting with artificial intelligence to deploying it at scale, the focus is shifting from standalone AI tools to agent-driven ecosystems that can execute tasks, make decisions and collaborate across business functions. Balakrishna DR (Bali), Executive Vice President, Global Services Head, AI and Industry Verticals at Infosys, speaks about the rise of AI agents, the importance of context engineering, governance challenges and how enterprises need to prepare for an AI-first future.
There is growing discussion around AI becoming the "new interface for everything". How realistic is that shift?
Conversational and agentic interfaces are rapidly becoming the default layer for many digital interactions because they reduce the distance between intent and execution. That is why most enterprise software vendors are putting AI agents in front of their products.
The change is already visible in customer experiences. Infosys, for instance, developed the Ask Ralph system for Ralph Lauren, which provides personalised outfit recommendations within the retailer's app. In enterprise settings, users can simply ask for a sales forecast and AI agents can gather data, analyse it and deliver results without requiring navigation across multiple systems.
However, graphical interfaces are not going away. Tasks that require precision and speed will continue to rely on traditional interfaces. The future will likely be a combination of natural-language interactions and graphical user interfaces, with the real value shifting towards context and orchestration.
What role will multi-agent systems and context engineering play in enterprise AI?
Multi-agent systems are already being deployed across business functions such as compliance, HR, procurement and finance. These agents increasingly work together to deliver business outcomes rather than perform isolated tasks. Infosys has built multi-agent workflows for clients such as Americana Restaurants to automate invoice processing, while supply-chain environments are seeing agents coordinate orders, inventory updates and document processing.
The critical differentiator is context engineering. Enterprise context goes far beyond a knowledge base and includes policies, business rules, historical decisions, permissions and real-time data. If that context is not managed properly, errors can multiply across agent workflows.
Take insurance claims processing. Claims, fraud, policy and customer-service agents can work together to settle claims far more quickly, but only if they share the right context. As enterprises move towards an "Internet of Agents", context engineering is becoming a foundational capability.
AI-native software engineering and 'vibe coding' are gaining attention. How will they reshape application development?
AI-native software engineering is compressing parts of the development lifecycle and allowing engineers to focus on higher-value problems such as legacy modernisation, performance optimisation and integration of fragmented systems.
The result is software that is more adaptive and aligned with business requirements. User experience becomes a primary design objective rather than an afterthought, leading to more responsive and frictionless applications.
At the same time, there is a risk of low-quality or hallucinated AI-generated code entering enterprise environments. If left unchecked, it could create outages and security issues. As a result, systems integrators will increasingly focus on building governed development environments that allow organisations to use AI safely at scale.
How important are governance and responsible AI frameworks as AI becomes more autonomous?
Governance becomes critical as AI systems gain greater autonomy. According to Infosys research, 95% of surveyed executives have experienced at least one AI-related incident, highlighting the growing risks.
Traditional guardrails designed for language models are not sufficient because AI agents can take actions, not just generate content. Enterprises need behavioural and intent-based controls aligned with business policies.
Balakrishna argues that enterprises should start treating AI agents like members of the workforce, with identity management, access controls, audits and performance reviews. Trust and responsible AI will increasingly become key competitive differentiators as customers gravitate towards systems they can rely on.
How is Infosys Topaz helping enterprises build AI systems, and what demand trends are you seeing?
Client demand is largely centred on six areas: AI strategy and engineering, data for AI, process AI, agentic legacy modernisation, physical AI and AI trust. Infosys Topaz is designed to help enterprises move beyond pilot projects by bringing together data, models, agents, workflows, governance frameworks and ecosystem partnerships. The goal is to provide the production-grade foundation required for large-scale AI adoption while ensuring security and business alignment.
What will enterprise interaction look like in an AI-first world?
Enterprise interaction will increasingly shift from people navigating software to AI agents navigating ecosystems on their behalf. Balakrishna believes this will lead to an "Internet of Agents", where autonomous systems collaborate across enterprises, suppliers and partners to execute transactions and business processes.
This could eventually create an agent economy, with organisations publishing specialised agents and capabilities through marketplaces, much like applications are distributed today.
The organisations best prepared for this transition are already treating AI agents as a core part of their operating model. They are investing in data foundations, modern architectures, workforce upskilling and governance frameworks. In the AI-first era, success will depend not just on building intelligent agents, but on creating trusted and scalable agent ecosystems.
Published by HT Digital Content Services with permission from TechCircle.