Rahul Kumar Bet His Career on Machines That Talk Back

Every product person eventually picks a hill. Rahul Kumar picked one that most people spent a decade laughing at: the idea that you could call a company, type a company, message a company, and a machine on the other end would actually understand you and help. For years that machine was a punchline. The chatbot that looped. The IVR that could not hear your accent. The "did you mean" that never did.

Then, quietly, it started to work. And Kumar, who had been standing in that exact spot the whole time, found himself holding a decade of hard-won knowledge about the one interface everyone suddenly wanted to build. His resume is almost a map of how customer service became the front door to enterprise AI: IBM in the early conversational era, Jio Platforms at the scale of a billion users, Talkdesk in the contact center, and now Druid AI, the agentic platform where he runs the business across India, APAC and the Middle East.

01The bet nobody wanted

It is easy to forget how unglamorous conversational AI was for most of its life. The models were brittle. The demos oversold. Enterprises bought a pilot, watched it stumble on real customers, and quietly shelved it. Selling into that skepticism is its own kind of craft - you are not selling a miracle, you are selling patience, integration work, and a roadmap that has to survive contact with a call center's worst Monday.

That is the work Kumar kept choosing. At IBM he learned the discipline of enterprise scale, the era when "conversational AI" mostly meant Watson-branded assistants and a lot of careful expectation-setting. At Jio Platforms he saw what happens when you point technology at a market measured in hundreds of millions of people, where the difference between a system that works and one that almost works is not a rounding error - it is tens of millions of frustrated humans. At Talkdesk he moved to the operational heart of the problem: the contact center, where every design choice shows up as a hold time, a resolution rate, or a customer who hangs up.

There is a version of the AI story that lives entirely in benchmarks and model releases. Kumar tells the other one. Ask him about the future and he does not start with parameters. He starts with data hygiene, consent capture, and fraud - the plumbing that decides whether any clever model survives its first week in production. It is the difference between someone who has watched a demo and someone who has watched a deployment.

02When the customer is a bot

Kumar's most interesting argument is also his most unsettling, and it comes straight out of his financial-services work at Talkdesk. For decades, the entire discipline of marketing and customer experience assumed one thing: on the other end of the interaction is a human being with attention to win, a mood to manage, a memory to build loyalty in. Kumar thinks that assumption is quietly breaking.

"Financial institutions are no longer just competing for customer attention; they're competing for visibility within AI-driven environments."

The reasoning is simple once you see it. AI agents are starting to do the shopping. They research, compare, and increasingly transact on a person's behalf. When that happens, your beautifully designed landing page is no longer being read by a person deciding whether to trust you. It is being parsed by a machine deciding whether you even qualify for the shortlist. Your best salesperson, in that world, is a clean, structured, machine-legible data feed.

Which leads to the line that should make a lot of brand managers uncomfortable: "If product data is unstructured or inconsistent, it may not surface correctly in AI-driven recommendations." Translated: messy data is invisible data. No visibility, no recommendation, no sale. Governance stops being a compliance chore and becomes a growth lever.

03Loyalty becomes a calculation

If agents can re-check the whole market every time they are asked, then loyalty stops being a warm feeling and starts being a running score. Kumar puts it plainly: customer loyalty is becoming more performance based, and consumers - or the agents acting for them - can continuously reassess options. Transparency, pricing and product clarity move from nice-to-have to the whole game.

This is the part that flips a lot of conventional strategy on its head. For years the CX playbook was about adding channels: put a bot on the website, add WhatsApp, add voice, add SMS. Kumar's advice runs the other direction. Stop optimizing channel by channel, he argues, and orchestrate the whole journey, because the AI on the other side does not care which door your customer walked through - it cares whether the experience holds together end to end.

"Move beyond channel-specific optimization and focus on end-to-end orchestration."

It is a deceptively big ask. End-to-end orchestration means your pricing engine, your product catalog, your consent records and your fraud checks all have to speak the same language and agree with each other in real time. That is not a chatbot project. That is a plumbing project dressed up as an AI project, which may be exactly why the people who understand it best are operators like Kumar rather than the researchers who get the magazine covers.

04The career as an argument

Read Kumar's path forward and it looks like a series of steadily larger bets on the same idea: that the conversation, not the app, becomes the interface. IBM gave him the grammar of enterprise software. Jio Platforms gave him scale most people never touch. Talkdesk put him inside the contact center, running the financial services and insurance business - an industry where a wrong answer is not an inconvenience, it is a regulatory event. Druid AI, where he now leads India, APAC and MEA, is the bet that agents - not just assistants - are the next layer.

Notice the geography, too. The interesting move is not a single leap but the direction of travel: from a global giant to India's biggest platform to an American scale-up to a European AI company, all while staying pointed at the same problem. That is rare. Most careers drift toward whatever is hot. Kumar's reads like someone who found the current early and kept swimming in it, picking up a different lens at each stop - governance at IBM, scale at Jio, operational reality at Talkdesk, autonomy at Druid. Put those lenses together and you get an unusually complete picture of why enterprise AI succeeds or fails in the real world, which is almost never about the model.

Druid AI itself is a useful backdrop for his thesis. Founded in 2018 and based in Bucharest, it builds agents that sit on top of an enterprise's existing systems - talking to people, pulling data, and triggering workflows in the tools a company already runs. It raised a $31 million Series C in 2025 and pitches itself on governance and composability rather than a single all-knowing model. That is squarely the world Kumar has been describing for years: not one giant brain, but many governed agents wired carefully into the plumbing.

05What you can steal from it

There is a career lesson buried in here that has nothing to do with AI. Pick a wave early. Ride it through the boring years when the technology embarrasses you and the demos fall over. Be standing there, with a decade of scar tissue, when it finally breaks. Conversational AI was Kumar's wave. He did not join it after it became obvious; he was already fluent when the rest of the market woke up.

And there is a practical one for anyone who sells anything online. The uncomfortable reframe is that you may no longer be optimizing for a human's attention span. You may be optimizing to be legible to a machine that never gets tired of comparing, never forgets a better price, and never rewards you for a clever tagline. Clean your data. Make your pricing honest and readable. Build the consent and fraud scaffolding before you point an agent at a customer's wallet. Boring, unglamorous, and - if Kumar is right - the difference between being recommended and being invisible.

The machines are finally talking back. The question Kumar keeps pressing is the one most companies have not answered: when the thing on the other end of the line can decide, compare and buy, are you ready to be understood by it?