At the launch of ConvoZen, Express Computer caught up with Akhil Gupta, Co-founder and CTO of NoBroker, to understand the evolution of their AI journey from Callzen.ai to ConvoZen. In this candid interaction, Gupta shares how NoBroker is pushing the boundaries of conversation intelligence through multilingual voice bots, the challenges of building AI for real-world Indian conversations, and how ConvoZen is empowering call centres rather than replacing them. He also outlines NoBroker’s broader tech-first strategy in the property space and hints at what’s next for the platform.
The last time we interacted was around 15-16 months ago during the launch of Callzen.ai. Now, after 60-65 months, you have introduced ConvoZen. Could you share insights into the journey so far and how it has evolved?
Since we last spoke, I think you’ll find it interesting to know why Callzen.ai has now become ConvoZen. It’s not two separate products, it’s actually an evolution, an upgrade of the original product we started with.
Our AI journey at NoBroker began because we run a large call centre, being a tech-first company, we handle all customer service through this setup, with 3,000–4,000 people speaking to customers daily. Around 2017, we realised that if we could transcribe these calls, it would help us gain insights much faster. We managed to build our own small language model (SLM) back in 2019–2020, which enabled us to do this effectively.
This turned out to be a massive problem to solve, multilingual transcription, especially in the Indian context, is extremely complex. Conversations often involve language switches, like English mixed with Hindi, Tamil, or Telugu, and at that time, there were no models capable of handling this with high accuracy. That’s how Callzen.ai was born—the idea was to extract the “Zen” or insight from calls.
However, as NoBroker evolved, we noticed a significant shift, customers were increasingly communicating through chats, emails, WhatsApp, and other platforms. So, the same intelligence we applied to calls had to extend to all forms of conversations. That’s why Callzen became ConvoZen—short for “Conversation Zen.” It now works across voice, chat, email, WhatsApp—any conversation channel.
Since our last interaction, the product’s capabilities have grown exponentially. We now support almost all Indic languages and have introduced virtual agents. These AI-powered voice bots can hold conversations just like a human, in the customer’s preferred language. ConvoZen has been around for about 12–15 months now, and the virtual agent or voice bot component came into shape roughly four months ago.
How have voice bots improved conversation intelligence and quality tracking? Could you also outline the language support offered across major Indian languages?
Voice bots have significantly enhanced conversation intelligence, quality tracking, and adherence. They help analyse conversations more effectively while taking over routine, repetitive tasks that don’t require human intelligence. This allows human agents to focus on more complex and value-driven interactions.
In terms of language support, we cover almost all major Indian languages—English, Hindi, Tamil, Telugu, Marathi, Gujarati, Malayalam—essentially all widely spoken regional languages across the country.
What major challenges did you encounter, particularly on the technical front?
One of the biggest challenges we faced, especially on the technical front, was language complexity. Building for English is relatively easier, but Indian languages are a different ball game altogether. Most foundation models or available APIs for transcription are designed to work in very controlled, studio-like environments. But real-world conversations, especially in call centres, are far from that.
There’s constant background noise, with multiple agents talking nearby, which makes transcription tricky. Also, these standard APIs assume the conversation happens in a single language, but in India, it’s common for people to switch between English and regional languages mid-sentence—like going from English to Kannada in the same breath.
Then there’s the challenge of dialects. The way languages like Hindi, Tamil, or Telugu are spoken can change every 100 kilometers. We had to train our core SLM (spoken language model) to handle all these nuances and still deliver high transcription accuracy. That was one of the toughest problems we had to solve.
Considering the property tech industry as a whole, how do you position the NoBroker platform in terms of its technological capabilities? While the no-brokerage model is a standout USP, how does your platform differentiate itself from competitors, particularly at a local level?
I think we’re positioned very differently in the proptech space. To begin with, we don’t really have a direct competitor—our main competition is actually the traditional broker, who doesn’t use any technology.
People often assume that classified platforms are our competitors, but they’re not. Those platforms simply connect people and step away. At NoBroker, the tech journey starts the moment someone registers. From the very beginning, we use AI to verify whether a user is a genuine owner, tenant, buyer, or seller—because we need to weed out brokers from the platform.
Our technology goes further—telling property owners how many days it will take to close their listing, helping tenants and buyers find the most suitable properties, estimating the closing price of properties, and even anticipating when someone might need services like cleaning, painting, or packers and movers.
In that sense, we’re a truly high-tech platform. I don’t think anyone comes close to what NoBroker is doing in terms of end-to-end technological innovation in this space.
Returning to the topic of ConvoZen, do you think implementing these technologies in traditional call centres, particularly by legacy companies, could negatively impact employee strength and the workforce?
No, I don’t think so. We need to understand that technological advancements will keep happening—it’s inevitable. Think back to the late ’80s when computers came in. Did they reduce the number of jobs, or did they help people do their jobs better? When the internet and dot-com era arrived, or when platforms like Zoom emerged, suddenly it became possible to do interviews from your home or office, even with someone sitting in San Francisco. Earlier, that would have meant a lot of travel and hassle. Now, you can do three to four interviews a day instead of one every few days—at a fraction of the cost and effort.
Similarly, AI is here to make people more efficient. I always say, if you don’t use AI in your job, your job may be at risk. But if you use AI to do your job, you can multiply your efficiency and output.
Systems like ConvoZen are not replacing jobs; they’re empowering agents. For call centre agents, ConvoZen equips them with the knowledge they need to serve customers faster, with less stress, and ultimately do more meaningful work. As for virtual agents or voice bots, they handle repetitive, mundane tasks that don’t require human intelligence—freeing up humans to focus on higher-value work.
So, when it comes to legacy companies implementing ConvoZen, it’s not about reducing workforce. It’s about hitting the bottom line positively—improving efficiency and enabling the same team to serve many more customers effectively.
During the launch event, Google Cloud India showcased its video capabilities. Are there any plans to leverage this technology for your customers or potential customers, particularly for property viewings?
I don’t think we have any immediate plans around that right now. What Google Cloud India showcased were mostly alpha-stage products. As and when those technologies mature—and if we see a relevant use case, we’ll definitely be among the first to adopt them.
What’s next for NoBroker? Could you share what new developments or innovations are on the horizon?
This is the big new thing we’ve launched, and given the pace at which we’re innovating and rolling out new offerings, our focus right now is on strengthening our core businesses NoBroker, NoBrokerHood, Home Services, and now ConvoZen. ConvoZen, in particular, is seeing strong demand because AI is clearly helping businesses deliver better customer service. We’re just getting started on that front, and there’s a lot more to come within ConvoZen itself. We’ll keep you posted as new developments unfold.