In a rapidly shifting digital landscape, where AI capabilities are evolving faster than organisations can absorb them, Harnath Babu, CIO of KPMG India, offers a grounded yet visionary perspective on what truly matters. In this candid conversation with Express Computer, he explains how enterprises have moved beyond experimentation and are now deploying complex, high-ROI AI use cases that are reshaping business performance.
Harnath highlights a striking trend: AI models that could solve only 3–5% of coding tasks three years ago can now handle nearly 80%, dramatically compressing development cycles — in some cases turning two-year projects into 40-hour builds. He also reveals how Agentic AI is emerging as the next architectural leap, where business processes are decomposed into autonomous, goal-driven agents that work in parallel, transforming workflows like market research, cybersecurity, and invoice processing.
Some edited excerpts:
We’ve heard a lot of AI use cases today—mostly simple ones. Let me begin by asking you: what are some complex AI use cases that actually deliver high ROI?
What I’ve observed today is a clear maturity in AI conversations. A year ago, many organizations were simply exploring what AI could do. Today, leaders are discussing industry-specific use cases, adoption levels, and, most importantly, ROI—whether it’s convincing the CFO or ensuring the business truly understands the value.
Within KPMG, our own journey mirrors this evolution. We moved from basic bot-generated responses to highly contextual, transaction-driven conversations powered by advanced AI models. Today, AI helps us prepare for client engagements with precision—pulling financial disclosures, analyzing media sentiment, identifying opportunities, and presenting sharp insights with a single prompt.
From an ROI perspective, this shift is massive—better quality conversations, greater efficiency, and significantly higher conversion rates in client engagements.
Across industries too—logistics, healthcare, cybersecurity—we are seeing real gains. One compelling example is cybersecurity automation. Agentic systems can now monitor multiple threat sources, correlate signals, and autonomously trigger actions aligned with SOC playbooks. The time to detect and respond has dramatically reduced.
The larger point is: complexity is increasing, maturity is increasing—and the speed at which organizations translate an idea into a deployed AI use case now defines competitive advantage.
You’ve mentioned Agentic AI multiple times. How do you see its promise and where do you see the biggest opportunities?
Think of Agentic AI like microservices. Earlier, applications were monolithic; then we broke them into microservices. Agentic systems break business processes in the same way.
You take a workflow—say market research—and split it: one agent fetches financials; one collects media sentiment;one pulls CRM data and one retrieves regulatory updates
An orchestration agent then analyses the combined inputs and produces an expert-level output. All this happens in parallel, not sequentially. That’s where the magic is.
Another example is invoice processing: An agent can extract invoice details, another pushes data to SAP, another triggers vendor reminders. Tasks that once required manual operators are now autonomous.
This is not science fiction. It is happening today.
And importantly—humans are not removed. They become supervisors who trigger, validate, and guide the agents.
Prediction is done by AI; judgment stays with people. That balance is crucial.
India is a services-driven economy with millions working in IT and operations. With the rise of such cognitive automation, how can India remain competitive? What skills must the workforce develop for this new era?
India’s strength has always been people. But traditionally, we’ve solved problems by adding more humans. That mindset must change.
Here’s a fact: Three years ago, code-generating models could solve about 3–5% of coding tasks. Today they solve 80%. That is staggering. A startup founder recently said what would have taken him two years to build, he shipped in 40 hours.
That is the new reality.
Spec-based development—where you simply pass an application specification to the model and it builds the app—is already here. Testing is becoming autonomous. Application UI is automatically generated. This is the new baseline.
So what happens to our workforce?
We must move up the value chain.
Not coding—problem solving. Not syntax—product thinking. Debugging may still matter, but creativity and design thinking will matter more.
I tell my kids: you don’t need to learn Python first. Learn how to think with no-code tools, learn prompt design, learn how systems work. The world is moving toward vibe coding—building by describing.
Upskilling is non-negotiable.
Finally, what is your advice to the CEOs? Technology is evolving almost too fast. How should they prepare their organizations for the future?
It’s true—humans are adaptable, but the pace of change today is overwhelming even for the best organizations.
Innovation is happening everywhere, simultaneously.
My advice: Start small. Experiment. Create excitement.
Evangelize AI internally. Don’t announce job cuts—announce productivity improvements. Show your people what is possible.
In KPMG, we conduct masterclasses to help employees build small utilities, demos, and automations on their own. When they see what AI can do for their day-to-day activities, adoption takes off.
Make the conversation persona-based. For example, with our sales teams, I mapped their entire day—from preparation to meetings to follow-ups—and showed how AI can make each step 10x more productive. That’s when it clicked for them.
Most employees only know “ChatGPT works.” Beyond that, they don’t know the possibilities. It’s the CEO’s job to open their eyes.
AI is not the next trend—it is the next foundational shift after the Internet.
The sooner leaders internalize that, the better positioned their organizations will be.
Any closing thoughts?
Just one: The art of the possible with AI is enormous. CEOs should focus on asking the right questions—not having all the answers. That mindset will define the next generation of successful organizations.