EY helps clients create long-term value for all stakeholders. Enabled by data and technology, our services and solutions provide trust through assurance and help clients transform, grow and operate.
At EY, our purpose is building a better working world. The insights and services we provide help to create long-term value for clients, people and society, and to build trust in the capital markets.
Why India is an ideal test bed for
Agentic AI innovation
Listen to our latest podcast on how Agentic AI is moving beyond traditional automation to reshape enterprise workflows, public-facing services and the broader digital ecosystem.
In the latest episode of EY.AI Unplugged Season 2: AI Agents of Impact, a part of the EY India Insights podcast series, Rohit Pandarkar, Partner, Technology Consulting at EY India, discusses how Agentic AI is moving beyond traditional automation to reshape enterprise workflows, public-facing services and the broader digital ecosystem. Rohit highlights that India offers ideal conditions for Agentic AI to scale because of its large population, strong Digital Public Infrastructure and adoption of mobile-first platforms. He explains how diverse languages, high digital interaction and innovation-friendly policies together make India a natural real-world test bed for AI experimentation and impact.
Key takeaways:
India’s Public Digital Infrastructure and large user base lend to scalable Agentic AI adoption.
Indian startups and enterprises are likely to start deeper leverage of Agentic AI in 2026.
The clearest value of Agentic AI is in workflow automation as it streamlines workflows, saving time, effort and enterprise cost.
Organizations in the future could be diamond-shaped instead of pyramid-shaped, with smaller teams at the bottom as AI agents perform those functions.
Role-based access control (RBAC) is necessary to define levels of access allowed for agents.
Strong access controls and safety testing are essential to prevent prompt‑driven cyberattacks.
With Agentic AI, people will treat cybersecurity as another step to be covered. But the accruable benefits, especially for enterprises, may have a direct impact on the bottom line, compared with peers who do not leverage Agentic AI.
Rohit Pandharkar
Partner, Technology Consulting, EY India
For your convenience, a full text transcript of this podcast is available on the link below:
Pallavi
Welcome to EY.AI Unplugged Season 2: AI Agents of Impact, a special series by EY India Insights. In this episode of ‘India’s digital test bed: Where AI meets impact,’ we explore how India’s vast digital ecosystem, diverse data and innovation-driven enterprises are shaping the future of AI adoption not just for efficiency, but for meaningful impact. Join us as we uncover how India is becoming a living laboratory for AI transformation, where technology, purpose and progress converge.
Our guest today is Rohit Pandharkar, Partner, Technology Consulting at EY India. Rohit brings over 18 years of experience in technology and digital transformation. He has been a key leader in helping businesses embrace emerging technologies, particularly in AI and data analytics, to drive sustainable growth and competitive advantage. Rohit works closely with clients to shape their technology strategy, leveraging his deep expertise in IT modernization and AI adoption to deliver high-impact solutions.
Thank you, Rohit, for joining us today. It is a pleasure to have you here.
Rohit Pandardkar
Thank you, Pallavi
Pallavi
Rohit, to begin with, could you help our listeners understand what Agentic AI means and how it differs from traditional AI systems?
Rohit Pandarkar
Agentic AI goes beyond traditional predictive analytics or machine learning. It is about autonomous and goal-driven intelligence. For example, there is a master agent that coordinates multiple worker agents, all working towards shared objectives. They operate with shared memory, learn continuously from feedback and can execute tasks autonomously, if permitted. This means that they can do recursive self-learning and that with every cycle, every interaction they grow smarter, faster and become more capable. It is like having an evolving digital workforce that keeps improving as they learn on the job.
Pallavi
Thank you, Rohit. What makes India such a unique test bed for Agentic AI?
Rohit Pandarkar
India offers ideal conditions for Agentic AI to mature by scaling 140 crore people that India has, digital public infrastructure like Aadhaar, ONDC and NPCI, payment gateway UPI and also innovation-friendly policies such as IndiaAI Mission that make India a real-world sandbox for experimentation on AI.
The diversity of users, including local dialects and languages and the digital interaction savviness of the Indian public (for example, the usage of WhatsApp, Instagram, YouTube and even the UPI apps) allows these AI agents to seamlessly integrate into the mobile apps that Indians are anyway using. For example, where you could ask an agent to figure out the menu for a football watch party with your friends in the evening.
It also places India on the world map, where a lot of agentic learning can happen because of our 1.4 billion population, from payments to identity and an account aggregator. India has a digital stack that already makes autonomous agents easy to leverage this existing digital public infrastructure and do tasks and actions of business importance. That is why I feel Indian startups and enterprises are highly likely to start leveraging Agentic AI in 2026 to do more with it.
Pallavi
Thank you, Rohit. Could you also cite some examples of how Agentic AI can create tangible impact on enterprises or citizen-facing workflows?
Rohit Pandarkar
The clearest value of Agentic AI comes from workflow automation. It reduces time, reduces the motion required, improves the actual activity being done and reduces the effort of multiple folks inside an enterprise and also reduces costs across business processes. Think of scenarios like issuing a replacement debit card, onboarding a new customer, or processing a reconciliation job overnight. In all of these, Agentic AI can do the job better than human beings, provided there is a small human oversight over the actual business decision.
Agentic AI can autonomously manage these tasks by interacting directly with enterprise systems like customer relationship management (CRM) software, inventory management tools, loan origination systems inside a bank and supply chain management systems inside a manufacturing organization.
It is the kind of automation that does not just speed things up, but also frees human teams to focus on higher-value work, while AI continuously learns from the interactions in business jobs. In addition to that, I feel that the organizations of tomorrow are going to look diamond-shaped instead of pyramid-shaped, which means there will be a relatively smaller team size at the bottom, with the majority of that being taken over by AI agents, or at least they will be working in a hybrid format where humans are able to do 10x the job. Perhaps the job that 10x what human beings previously were doing with the help of all these agents, because they are digital workers, available at the call of human agents.
Pallavi
Thank you, Rohit. Lastly, as organizations scale these systems, how can we ensure Agentic AI is governed responsibly and aligned with trust frameworks?
Rohit Pandarkar
I would like to talk about the recent RSA Conference, which is the RSA encryption algorithm-related conference that happens for cryptography and encryption in the US every year. In the recent conference, there was a lot of talk about treating AI autonomous agents as non-human entities who are digital workers, just like human workers or employees inside an organization and they would have to be given role- based access control.
This role-based access control (RBAC) is necessary for these agents because we have to define what access they get, what access they do not get. Do they have access to salary data of employees or appraisal data of employees, or the pricing information, which is very sensitive and also the information about sales pipeline or the PII (Personally Identifiable Information) data of customers? Either you can allow or disallow these access controls, but at the same time, you also have to test the AI bots for hallucination, harmfulness, bias and fairness, as well as things like accountability and explainability, which means that if the AI agent is doing something, can you explain why it did so and who is accountable for that decision? No department head or employee who is human can ever say that “I do not know why this decision was taken because AI has taken it”. Ultimately, there has to be human accountability behind every decision made.
Finally, it has to be tested for new types of cyberattacks that are happening with Agentic AI, like prompt- injection attacks or prompt-leakage attacks, where someone can insert or inject a prompt into the system prompt or the mother prompt of the agent, making the agent foolishly believe that the instructions given by the attacker are true and the agent starts behaving differently or maliciously.
Secondly, a prompt leakage attack would mean that you just do a brute-force simulation of multiple interactions with the same bot so that, over time, the answers given, or with slight variations of questions and seeing how the answer changes, you can reverse engineer what prompt or policy has been set behind the bot. So, all such attacks have to be prevented by testing the bots against them.
I believe the arrival of Agentic AI raises further questions and security measures along with the benefits it gives. But at the same time, the benefits are likely to far outweigh the challenges we will face because cyberattacks have always been around even in the cloud age and the mobile age and ever since the internet arrived.
With Agentic AI, people will treat it as another step of cybersecurity that has to be covered. But the benefits that will accrue are for all of us to reap and especially for enterprises that may have direct impact on the bottom line, compared with those organizations that are competitive peers who do not leverage such things.
So, imagine that in six months, an Agentic AI system in an enterprise, irrespective of the sector that it belongs to, becomes so intelligent and works like a clockwork-style engine and delivers business results one day, one second, one job at a time. They are likely to run ahead of the peers in the competition because AI and robotic automation through the intelligence of GenAI is going to simply outweigh what human beings can do in this calendar year itself.
Pallavi
Thank you, Rohit. That brings us to the end of this episode. Thank you so much for joining us today and sharing your valuable perspectives with our listeners.
Rohit Pandarkar
Thanks, Pallavi, for hosting me.
Pallavi
Thank you to all our listeners. You have been listening to EY.AI Unplugged – Season 2: AI Agents of Impact, on the EY India Insights podcast. To explore more insights, reports and conversations on how AI is redefining business and society, please do visit our website, ey.com/in.
Thank you for tuning in and stay with us as we continue to spotlight the leaders and ideas driving AI with impact.
