How Google Backed Slang Labs’ AI Voice Assistant Is Changing The Way We Interact With Apps

Recent advancements in conversational AI have made automated voice agents smarter and more capable of accurately understanding sentiments and swiftly resolving customer queries. 

This has prompted several industries and sectors, including education, ecommerce, healthcare, banking, and insurance, to improve their customer support services using voice assistants.

As a result, the global voice assistant market is projected to surpass a market valuation of $38.5 Bn by 2031, growing at a CAGR of 26.45% from $4.6 Bn in 2022, as per Astute Analytica report.

Amid the current scheme of things, Google-backed Slang Labs is making waves in the conversational AI space in India.

Founded in 2017 by serial entrepreneurs Satish Chandra Gupta, Giridhar Murthy, and Kumar Rangarajan, the startup aims to reduce thought-to-action latency by using the power of voice inside apps. 

Simply put, through its conversational in-app voice assistant platform (CONVA.ai), Slang Labs helps brands make their apps faster, easier and more accessible by enabling multilingual voice-based interactions on their respective platforms.

“Our platform offers an in-app specific assistant, distinct from phone-level assistants like Google Assistant or Alexa. It boasts three key features — prioritising in-app functionality, multimodal interaction (voice and text), and multilingual support. Our voice assistant allows users to interact in their native language within English apps,” cofounder Rangarajan told Inc42.

In the conversational AI space, the startup competes with the likes of Observe.AI, Senseforth.ai, Yellow.ai, and ConveGenius, among others. The startup has raised a total funding of $2.39 Mn since its inception.

Slang Labs And The Three Musketeers

The story of Slang Labs began when the trio of Gupta, Murthy, and Rangarajan decided to work together again in 2017. The cofounders had last worked together a decade ago at Rational Software (a part of IBM).

However, this time, there was something quite massive on the agenda — taking the entrepreneurial leap of faith. The trio then got together on the drawing board to redefine the way users interact with apps, by clicking and typing. 

This was when they realised that the space was crying for disruption, majorly on the back of the English-only experience of a majority of apps. In pursuing this, the cofounders also decided to handle complexities that come with the rigidity of a touch-only experience like typing on a small keyboard or spelling mistakes.  

With a vision to enable users to communicate naturally with apps, with zilch language barriers, the cofounders incorporated Slang Labs. Today, the startup helps upgrade the potential of their clients’ apps by making them smart enough to understand and respond to users’ queries in a number of languages, including Hindi, Kannada, Tamil, Malayalam, Spanish, and Vietnamese.

Today, the startup specialises in providing multilingual in-app voice assistant technology to ecommerce players and brands with native online stores. 

“Given that most of us spend a significant amount of time inside ecommerce apps like BigBasket, Tata Neu, or Amazon, navigating through various sections and products. Typically, the interaction with these apps starts with a search, where we are seeking specific information or products,” Rangarajan said.

The cofounder told Inc42 that apps like Tata Neu, Nykaa, ICICI Direct, Redbus, SpiceJet, BigBasket, and Unilever, among others, leverage its AI platform CONVA.ai to make their user experiences seamless while interacting with their respective apps.   

For instance, instead of manually typing a search query for an air conditioner, users can simply ask, “What AC should I buy for my 10×10 room?”  

Besides, Slang Labs offers a search copilot feature, which assists users with three main types of searches in apps — navigation, information, and transactions. This helps users quickly find what they need, whether it’s checking statements in banking apps or  making purchases in ecommerce platforms. 

What’s Inside Slang Labs’ Tech Arsenal

Speaking about the technology stacks, Rangarajan said that Slang Labs uses various technology stacks, including Google’s speech-to-text engine, known for its quality. However, he mentioned that Google’s engine is generic and not optimised for specific domains. To address this, Slang Labs applies their algorithms to tailor it for specific purposes.

He added that CONVA.ai has not developed a proprietary LLM but relies on existing open-source LLMs that are fine-tuned for the system. 

Moving on, CONVA.ai, a Gen AI-powered multilingual copilot builder platform for apps, provides an “in-app salesperson” experience, which is capable of simulating offline store interactions for app users in their preferred language. Designed for multilingual, multimodal, and multi-platform functionalities, it serves brands in sectors including retail, ecommerce, banking, fintech, and insurance.

As a cloud-based full-stack solution, CONVA.ai is integrated into apps as a software development kit (SDK). Once integrated, the SDK enables voice search and other conversational features within the applications.

“Depending on the use cases which you want to enable inside your mobile or web app, the CONVA.ai SDK will need different amounts of effort and time to integrate. Our basic voice search implementation can be completed within a matter of minutes without our team’s help,” Rangarajan added.

The startup launched the upgraded version of its AI solution, CONVA.ai 2.0, in May 2023. While the original version enabled multilingual voice search experiences and was built on old-school NLP techniques, the current version of CONVA.ai is built on top of GenAI and enables users to give more broad and natural commands and conversational searches in their natural language.

CONVA.ai 2.0 addresses four key challenges — increasing average cart value by facilitating natural product additions, seamless integration with existing text-based search, reducing drop-offs, and enhancing overall customer experience. 

The startup’s new platform focusses on grounded conversation voice search, improving understanding of user intent alongside existing keyword-based engines. Currently, the startup is working on the third version of CONVA.ai.

Moving on, the conversational startup anticipates adding six more brands to its roster, including TATA Digital and Aditya Birla, in the coming months.

The startup launched CONVA Magic in December 2023. The Magic app enables users to experience CONVA.ai within ecommerce apps such as Amazon without Amazon directly integrating CONVA.ai into their app.

The Magic app can be utilised by installing it from the Android Play Store and selecting one of the 10 ecommerce apps on the home screen.  

Initial Hiccups & The Way Forward

Backed by Google, Endiya Partners, and 100X Entrepreneur, Slang Labs had to face multiple challenges at the outset. The concept of deploying their voice assistant within apps prompted many at the time. 

“Educating the market about in-app assistants versus phone-level assistants was the first challenge. Another challenge was the uncertainty about user interest in voice-powered experiences. Over time, as natural language processing technology advanced, adoption rates increased substantially,” said Rangarajan.

Technical issues were also a concern for the startup, as earlier systems struggled with reliability and understanding natural language. However, with advancements like GenAI, handling queries across different domains became quicker and smarter. 

“Earlier, moving from one domain to another required significant engineering effort. However, with advancements, we are able to cover multiple domains. The goal now is to smoothly integrate AI into users’ existing activities across various apps, offering a superior experience compared to traditional chatbots,” Rangarajan said.

Further, the startup plans to launch a self-serve version of their platform next month, to make it easy for mobile developers to easily add GenAI experiences inside their apps.

The company is presently exploring various revenue models, one of which is Monthly Slang Users (MSU). Under this model, customers are charged based on the number of end-users utilising the copilot feature within their app, with a fixed monthly rate per user. This allows the customers to have a more predictable pricing model, as per the startup.

“We have witnessed a 3x growth in revenue since last year, after moving to the GenAI version of our platform. We expect to continue this trend and grow at the same or faster rate this year,” Rangarajan told Inc42.

Going forward, the startup, in the short term, aims to focus on expanding language support, broadening domain coverage, and entering international markets.