'Agentic AI unlocks new business models for IT firms': Deloitte's Ashvin Vellody

Bengaluru: Agentic AI helps businesses move faster by reducing the time it takes to understand information and make decisions, said Ashvin Vellody, Partner, Deloitte India. Also, for IT firms, agentic AI unlocks new business models. In an interaction with Uma Kannan of DH, Vellody, whose focus areas include consulting companies across industries reimagine digital experiences, shape digital strategy and embed AI in their enterprise, spoke about GenAI and challenges that organisations face in implementing agentic AI. His current areas of interest are in Generative and Agentic AI. Edited excerpts:

What kind of role can agentic AI play in the growth of companies, especially IT companies?

Agentic AI is redefining how companies—especially IT services firms—approach growth and transformation. Rather than focusing on isolated use cases, the real opportunity lies in solving classes of problems by ‘agentising’ entire process flows. This involves stitching together multiple single-purpose agents that collaborate to deliver outcomes across a workflow.

For IT firms, this unlocks new business models centred around IP-led, product-driven services. It also reimagines how services are sold—moving from effort-based pricing to outcome-based engagements—and transforms delivery by enabling humans and agents to work as a team. These agents accelerate speed-to-market, reduce duplication, and improve win rates by up to 30%. For example, a global manufacturer has deployed agents to detect shipment distress and autonomously handle support tickets, while a consumer enterprise has streamlined procurement through multi-agent orchestration.

Are there any challenges in implementing agentic AI?

Yes, there are several practical challenges that companies must navigate when implementing agentic AI. The first is data—many organisations don’t have clean, structured, or accessible data to support agent workflows. The second is integration. Most enterprises have legacy systems, fragmented APIs (Application Programming Interface), and siloed platforms, which make it difficult for agents to operate end-to-end. The third is security and governance. Since agents can take autonomous actions, companies need strong controls to prevent errors, hallucinations, or compliance risks. There’s also a softer challenge: change management.

Business teams often hesitate to adopt agentic AI because the value isn’t immediately visible, and the shift in roles can feel disruptive.

What differentiates agentic AI from traditional automation and GenAI implementations?

The real difference lies in how agentic AI is designed to solve business problems. Traditional automation focuses on repetitive tasks. GenAI is great for generating content or summarising information. Agentic AI goes a step further—it’s built to solve entire workflows by stitching together multiple agents, each with a specific purpose. These agents don’t just respond; they plan, reason, and act. For example, in a customer onboarding process, one agent might validate documents, another might update systems, and a third might trigger follow-ups. Together, they deliver a complete outcome. This modular approach allows companies to reuse agents across different processes and scale faster. It’s not about building one-off bots—it’s about creating a system of agents that can work together, adapt to business logic, and deliver measurable results. That’s what makes agentic AI a game-changer for enterprise transformation.

How can businesses accelerate outcomes and compress time-to-value across functions like marketing, finance and customer service?

Agentic AI helps businesses move faster by reducing the time it takes to understand information and make decisions. When agents are deployed across a process, they can run continuously, handle tasks in parallel, and eliminate manual bottlenecks. This leads to faster execution and quicker realisation of value. For example, in finance, agents can handle reconciliations, reviews, and reporting with minimal human intervention. In marketing, agents can automate campaign QA and performance tracking. A global enterprise used agentic AI to streamline procurement by deploying multiple agents that worked together across sourcing, approvals, and vendor management. The secret is to build a central library of agents that can be reused across functions, and to set up cross-functional teams that can rapidly prototype and deploy solutions. When done right, agentic AI creates a multiplier effect across the business.

What challenges do organisations face when integrating agentic AI without causing fragmentation or bottlenecks?

The biggest risk in scaling agentic AI is fragmentation. If different teams build agents in silos, without a common architecture or governance model, it leads to duplication, integration issues, and inconsistent performance. Companies need a central design authority to guide how agents are built, deployed, and managed. This includes standardising tools, platforms, and DevOps practices. Without it, agents may not talk to each other, or worse, they may conflict. A global consumer company avoided this by creating a centralised platform and a library of reusable agents, which allowed different business units to plug into a common framework. Another challenge is prioritisation—without a value hub to track ROI and align efforts, teams may chase low-impact use cases. The solution is to treat agentic AI as a horizontal capability, with clear governance, shared infrastructure, and a focus on solving business-critical problems.