As enterprises look to extract real, scalable value from their AI investments, the limitations of one-size-fits-all models are becoming clear. Companies are evolving from single-model AI deployments to hybrid, multi-paradigm architectures—also known as Poly AI—and this evolution is laying the foundation for agentic AI. Infosys has launched over 200 enterprise AI agents that are designed to transform enterprise processes. In this interview, Balakrishna DR, executive vice-president, global services head, AI & Industry Verticals, Infosys, speaks to Sudhir Chowdhary on the market transition and how Poly AI architectures are enabling the rise of agentic AI. Excerpts:
Why is a hybrid AI model better suited for solving enterprise-scale challenges?
A hybrid AI model strategically combines multiple AI approaches—traditional machine learning (ML), large language models (LLMs), small language models (SLMs), distilled models, and domain-specific fine-tuned models—to tackle complex business problems more effectively than any single solution. This approach is like assembling a specialised team, rather than relying on one individual for everything.
For enterprise challenges, this method is particularly potent because businesses deal with diverse data types and requirements simultaneously. A hybrid system might leverage traditional ML for structured financial data analysis, LLMs for complex reasoning tasks, SLMs for cost-efficient edge processing, distilled models for faster versions, and fine-tuned models specialised for industry-specific terminology and workflows (e.g., healthcare diagnostics or legal document analysis). This combination delivers superior accuracy, reduces costs compared to exclusively using large models, and ensures data privacy through flexible deployment options on premises or in the cloud.
How is Infosys using hybrid AI models for business transformation?
Infosys employs a comprehensive hybrid AI strategy, integrating proprietary small language models with established large language models and traditional ML techniques. Our approach balances reliability, cost, and data compliance with a ‘Modelling-as-a-Service’ offering to help customers create custom models.
Infosys has developed industry-specific small language models focused on domains like banking, ITOps, and cybersecurity. They use local deployment models for customers preferring on-premise data processing, while simultaneously using large language models for reasoning tasks. Traditional ML models remain central for many business processes involving transactional data.
How is Poly AI helping in the shift towards agentic AI systems?
The transition to agentic AI systems capable of independent planning, execution, and adaptation requires a foundation of reliable, interconnected AI components. Hybrid AI provides this foundation by creating robust systems that can handle the multi-step reasoning and diverse data processing autonomous agents demand.
Agentic systems operate by creating execution plans and employing specialised tools to interact with enterprise systems. This is where hybrid AI becomes crucial. Each tool interaction may necessitate different AI capabilities: domain-specific fine-tuned models for accurate interpretation, distilled models for rapid decision-making, and verifier models that act as quality checks to mitigate agent mistakes.
For example, when an agentic system processes a complex financial transaction, it might use a fine-tuned banking model for regulatory requirements, a distilled model for quick document classification, a reinforcement learning (RL) model to optimise validation steps, and a verifier model to ensure compliance. This hybrid approach guarantees both accuracy and reliability essential for autonomous systems handling critical business processes.
Any examples of how agentic AI is being used in different industries?
Agentic AI is already being deployed across various industries and verticals, particularly in business, IT, and operations. For a major consumer tech organisation, a deep research agent was developed to assist teams in efficiently researching complex technical questions. This resulted in substantial and consistent improvements in productivity and customer experience. Similarly, for a technology provider, agents have been deployed to transform the financial accounting process, directly impacting business outcomes and employee productivity. The adoption of agentic AI engagements is steadily increasing across Infosys’ customer base.