The enterprises that treat AI as another efficiency tool will capture a fraction of its value. The ones that use it to improve the quality of their decisions will redefine their industries.
Raj, you’ve built technology companies through several major shifts, from enterprise IT and mobility to conversational and agentic AI. What early experience still shapes how you think about helping enterprises adopt new technology?
In the 1990s, I built Intelligroup around SAP implementations for large enterprises. I learned that the scale, security, compliance, and enterprise integrations are non-negotiables. What separated successful projects was a business leader who owned the result and was willing to redesign the processes around technology.
I saw the same pattern again in the mobile era at Kony. The companies that redesigned their customer journeys around the phone got real returns. Those that moved old web forms onto a smaller screen saw little change. Today, AI agents are going through the same test, and the companies that rethink the work around them will move ahead quickly.
That experience shapes how we run Kore.ai. We start every engagement with the business result the customer wants, and we prove it on their own data before they commit. The technology disruption happens every few years. The way enterprises adopt it has changed very little in thirty years.
You founded Kore.ai in 2014, well before generative AI became mainstream. What did you see about enterprise technology then that made you believe AI would eventually become much more than a conversational interface?
When we started Kore.ai in 2014, people already spent most of their day in messaging apps. Enterprise software, however, still made them navigate complex menus and forms to get simple things done. I believed that conversation is natural to us and it would become the main way people work with enterprise systems; no UI is the best UI.
Conversation was only the starting point. Once a system understands what a person wants, the next step is to do the work for them. That requires secure connections to systems of record and a clear understanding of business rules, so we pioneered the conversational intelligence capabilities using NLP/NLU in our platform from the start.
Large language models came much later and made the understanding part far better. By then, our integration and governance work was already running in production at large enterprises. For us, agents became just the next chapter of a plan we started ten years ago.
Also Read: AITech365 Interview with John Riewerts, Chief Product & Technology Officer of Acoustic
Kore.ai has evolved from conversational AI toward agents that can reason, coordinate, and execute work across enterprise workflows. What changes when AI moves from answering questions to actually executing work?
When AI answers a question, a wrong answer costs a few seconds and some trust. When AI executes work, a wrong action can move money or change a customer record. The bar for reliability rises sharply, and so does the need to explain every decision.
Our 2026 Agent Productivity Index found that 79% of enterprises have reversed an action taken by an AI agent, and 70% have faced a failure their teams could not trace. Agents that execute work need a dual-brain architecture. One side reasons through complex situations and decides how to proceed. The other enforces a deterministic flow for business rules, compliance and guardrails, and records an audit trail for every action.
The role of people changes too. As AI takes on more of the execution, employees become coaches and supervisors who guide agents, review their performance and decide where they need to improve. At a leading bank in the Middle East, AI now handles memo parsing and data validation for credit evaluation, and deal evaluation dropped from days to hours. The credit officers trust the system because every step has an audit trail and they approve every final decision. The enterprise stays accountable for what its agents do, and its people spend their time on outcomes instead of routine tasks.
McKinsey’s 2026 research found that 80% of respondents report AI improving individual productivity, while only 37% report a positive impact on enterprise EBIT. What separates organizations that turn productivity gains into measurable business value?
That’s absolutely true, and I keep saying that a productivity gain at the individual level does not necessarily translate into business gain unless it’s directly tied to business metrics.
The business impact due to productivity gains on customer support usecases is directly measurable. AI-powered call summarization, for example, saves 10 seconds per call, which means millions of dollars in savings for the company, more calls handled, and standardized call summaries captured that help in call quality assessment and training needs.
EBIT moves when AI owns a measurable piece of work end to end, such as a resolved ticket or a processed claim.
The second thing successful companies do is keep improving Agents after go-live. Most agents decay after launch as models drift and the business changes underneath them. The gains turn into lasting EBIT impact when every agent is measured and tuned continuously against a business metric.
As enterprises move from individual AI tools toward agents embedded across workflows, the technology begins changing how work itself is organized. What do leaders need to rethink before moving from AI adoption to an AI-native operating model?
Most organizations add AI to processes that were built around human handoffs and approvals. An AI-native model starts with the outcome and decides which steps an agent owns and which steps need a person. Leaders need to redesign the work before they deploy the agents.
We made this change in our own platform engineering. AI agents now produce about 6,500 commits a month on our 2.6 million-line codebase, governed by 68 always-on guardrails. Our engineers spend more of their time on design and review, and we had to change how we measure their contribution. Our customers see the same result when they use AI to build, govern and optimize their AI agents. Engineering effort drops, so they can expand AI across more of the business without growing their teams at the same rate. That is what makes the AI-native operating model significant.
Ownership is the other big shift. Every agent needs a business owner accountable for its results, the same way a manager is accountable for a team. Leaders who settle ownership early will scale much faster than leaders who leave it to IT.
You’ve spoken about AI sprawl as organizations adopt agents across teams, platforms, and frameworks. Where should enterprises establish stronger central control, and where should they preserve the freedom for teams to experiment?
Agent development is now democratized. Anyone in the business can build an AI agent without waiting for IT, and that speed is valuable. It has also created agent sprawl and left engineering teams to track, control and maintain agents they never built.
Most of these agents run on different technologies and follow no common standards. Many of their builders have never shipped software to production. One CISO told me his team found three separate lending agents during a routine audit. None of the three teams knew the others existed, and the company had no view of the whole picture.
My rule is simple: decentralize the building and centralize the standards. Teams should pick their own use cases and experiment quickly. The enterprise should control what creates risk when it diverges: identity, data access, policy enforcement, audit, and one registry of every agent in production. We built ABL, our Agent Blueprint Language, and Arch, our AI architect, for this problem. A business user describes the agent in plain language, and Arch builds it in ABL with the right architecture and guardrails in place. The business keeps its speed, and engineering gets agents it can read, govern and maintain.
Kore.ai has taken an open approach across models, clouds, and enterprise systems. After decades of building enterprise technology, what have you learned about when a platform should make strong choices and when flexibility creates more lasting value?
We always believed in offering flexibility to choose what’s best for their enterprise. A platform should make strong choices about the things an enterprise should not have to solve repeatedly, such as security, governance and reliability. Those capabilities need to be consistent regardless of what technology sits underneath.
That covers how agents are defined, tested, governed and audited. We built ABL, our agent blueprint language, so every agent is machine-readable and inspectable in the same way, regardless of who built it.
Flexibility matters where enterprises need options to choose best-of-breed or when the market changes fast, like models today. So we stay open on models, clouds and enterprise systems, and we stay firm on the harness layer that makes agents reliable. In thirty years, I have seen platforms that over-control lose customers and platforms that under-control lose trust. The lasting value sits in getting that balance right.
Customer experience remains a major part of Kore.ai’s business, while agents are moving beyond conversations into actions across entire customer journeys. What changes when AI can resolve a customer issue rather than simply respond to it?
I think it changes what we should expect from customer service. For years, the industry has worked on individual touchpoints, making calls shorter and answers more accurate. But for customers, they care about whether their needs were served.
AI Agents can go past the basic interaction. An agent can understand what the customer is trying to get done, work across the systems involved and complete the steps to resolve it. A healthcare provider in North America for example, has AI agents book and update patient appointments end to end, inside their scheduling systems. Those agents contain 24% of interactions, a number that looks modest on a CX dashboard, but they have also enabled $3.2 million in revenue and a 468% return on the program. The focus moves from handling the contact to completing the work behind it.
The next step is preemptive service. When you already know something has gone wrong and you have the context to fix it, there is no reason to wait for the customer to find out and call you. More service will happen before the customer ever has to ask. For me, that is where AI will change customer experience the most.
You’ve argued that India should measure how much AI it creates, not simply how much it consumes. Having built a global AI company with a major R&D presence in India, what should India become known for next?
India is one of the largest users of AI tools like ChatGPT. I want India to become the largest creator of AI solutions. A generation of Indian engineers used programming languages to build software for the world. This generation should use AI infrastructure the same way, and take AI creation to the grassroots.
Most of India’s technical talent today works in technology companies in a few large cities. The bigger opportunity is a student in a village who builds with AI, or a farmer or teacher who solves a problem in their own community. The building blocks are in place. Multilingual models work across India’s 22 official languages, and the IndiaAI Mission makes compute affordable. India needs to start in schools and make AI building skills available everywhere.
That is why I have argued that India should measure how much AI it creates, alongside how much it consumes. When hundreds of millions of Indians become creators, they will solve problems unique to India, and many of those solutions will work for the rest of the world. That is the bigger opportunity.
After building companies through several technology cycles, what is one belief about enterprise technology that you once held strongly but eventually had to unlearn?
My core belief has not changed. In fact, it has grown stronger. I have always believed in using technology to solve enterprise problems at scale, and every company I have built rests on that idea. That held true through ERP, the internet, mobile and now AI.
What I have revised is my view of what technology can do. For thirty years, every major shift helped enterprises work faster and simpler. ERP standardized processes, and mobile put them in people’s pockets. Each one improved how we execute work. AI is the first technology I have seen that improves how we think. It helps people reason through a problem, weigh options and reach better decisions.
That makes this the largest disruption I have seen in my career, and it will touch every industry and every function. The enterprises that treat AI as another efficiency tool will capture a fraction of its value. The ones that use it to improve the quality of their decisions will redefine their industries.
Thanks Raj!
About Raj Koneru,
Raj Koneru is the Founder and Chief Executive Officer of Kore.ai, an enterprise AI company specializing in conversational AI, generative AI, and agentic AI solutions. With decades of experience in technology and entrepreneurship, he has led businesses through major industry shifts, including enterprise IT, mobility, and artificial intelligence. Since founding Kore.ai in 2014, Raj has focused on helping organizations improve customer experience, enhance employee productivity, and automate business processes through AI-powered technologies. Under his leadership, the company has expanded from conversational AI into enterprise AI platforms that enable businesses to build, deploy, manage, and govern AI agents across workflows.
