How NoBroker’s in-built AI model Convozen leverages insights from customer interactions

When one thinks of renting a house in any major city in India—without having to go through brokers—proptech unicorn NoBroker comes to mind. As its name goes, the company eliminates the middleman and connects tenants directly with owners for real estate deals. 

Established in 2014 by IIT alumni Akhil Gupta, Amit Kumar Agarwal, and Saurabh Garg, NoBroker offers a host of real estate solutions, including rental, sale, or resale of properties, loans, packers and movers, legal documentation, etc. 

In November 2021, the startup became India's first proptech unicorn—companies with valuations over a billion dollars—after raising $210 million. 

“We wanted to enable a platform using technology, where owners and buyers could sell and communicate without a third party who’s trying to push the deal for their benefit,” says Akhil Gupta, Co-founder and Chief Tech and Product Officer of NoBroker.

However, for the past seven years, the company has been dabbling with artificial intelligence (AI); it does not want to miss out on the opportunity. 

NoBroker’s latest business-to-business (B2B) offering, Convozen, is a small language model (SLM) that offers service agents tools to enhance performance and customer satisfaction. 

To the uninitiated, SLMs are AI models that use a smaller neural network, fewer parameters, and training data compared to an LLM. 

The AI-driven conversational intelligence platform provides insights from customer conversations to improve a company’s overall business efficiency. It uses advanced machine learning (ML) models to transcribe and extract insights from conversations conducted in multiple languages.

Product offerings

Oftentimes, manually auditing data emerging out of customer interaction is a monumental task for businesses. Gupta saw similar challenges while carrying out NoBroker’s daily operations, which became the stepping stone for building Convozen. 

“We are a technology play; we don't have feet on the street. All of it is done using technology. With loads of conversational data in hand, manual audits became challenging. However, human intervention is still essential as it's a high-value transaction," the co-founder explains.

The B2B offering, Convozen, integrates standard operating procedures (SOPs) and a knowledge base within its platform, streamlining business operations.

The platform leverages AI to evaluate consumer interactions across channels, ensuring adherence to SOPs and providing automated quality assessment. It uses ML to transcribe and extract insights from these conversations accurately. 

Convozen supports multiple channels—including chats, emails, tickets, and social media platforms—where consumers may cite their concerns. 

"It's a multilingual AI platform that transcribes all customer-agent conversations. With such extensive conversations, we want to find the ‘zen’ out of each interaction,” Gupta says.

At present, Convozen transcribes six languages, including English, Tamil, Telugu, Kannada, Hindi, and Marathi. “Other three languages—Gujarati, Bengali, and Malayalam—are currently undergoing private beta testing," he highlights.

Also Read

How NoBroker created a niche in real estate space by eliminating brokerage

Extracting insights

In 2019, Conzoven started transcribing customer calls to resolve the issue of good transcription solutions for call centre conversations, especially with regional languages and noisy calls. According to Gupta, Convozen records 8,000 hours of calls daily.

Its Pitch Pop feature summarises previous conversations and customer mood for service agents to make effective pitches in future calls. On the other hand, the Agent Assist feature offers real-time support and guidance to service agents during customer interaction.

These features compile conversation history into summaries, giving agents a better understanding of customers. The platform provides a personalised dashboard for agents, showing previous queries, action items, and talking points to help them improve their service and communicate more effectively with customers.

Gupta explains, "[The platform] semantically tags key talking points in conversations by understanding the intent and assigns these tags to the appropriate speakers. This allows for clearer tracking of who said what and the main topics discussed during interactions, further training the agents to perform better."

Besides this, Convozen has a mechanism for companies to rate the customer-agent interactions, which NoBroker calls ‘sentiment analysis’, by marking them either positive or negative. This process involves monitoring conversations continuously to assess the quality of interactions between customers and agents.

In fact, the platform lets businesses know where their agents lack through the agent performance tool, which monitors various metrics and provides customised training to improve their work.

Convozen also supports API connections of major CRMs like SalesForce and Zoho for easy integration.

Impact on business performance

Convozen captures key points and actionable insights from every customer interaction. It claims that business performance sees significant improvements, with up to 90% of the call audits automated and a 30% enhancement in agent performance.

According to a Business Research Insights report, the global conversation intelligence software market is expected to reach $2251.8 million by 2029, growing at a CAGR of 12.1%. 

Within just three months of its launch, the company onboarded over 15 customers in the edtech, healthcare, and fintech sectors, including companies like Lendingkartand Tata AIG. 

NoBroker has declined to disclose its subscription fees.

Some of the existing players in the market include Mihup, CallMiner, Convin, and Observe.

"Our tool, coupled with these transcripts, has better transcription accuracy compared to other tools, with an error rate ranging between 8% and 12%,” the co-founder says. 

This, Gupta adds, caught Google's attention, which invested $5 million in the company in an extended Series E round in 2023. 

Today, Convozen tracks over 97% of conversations, providing filterable and reportable data, groups them and generates comprehensive reports for analysis. 

Gupta believes as the demand for conversational intelligence software grows, handling large volumes of customer interactions while maintaining high levels of reliability may become a challenge. 

"We're currently developing a voice bot model to automate repetitive tasks using AI. Our goal is to enable our agents to focus on tasks that require human intellect, rather than repetitive ones," Gupta says.