In Hindi and Urdu, “gupshup” (गपशप) refers to informal chat or gossip. It is a fitting name for a company that provides a mobile-first, conversational messaging platform in emerging markets. Gupshup serves more than 10 billion messages a month and counts as its customers Google, Facebook (now Meta), Walmart, Twitter (now X), LinkedIn, Netflix, Visa and many other big companies.
The AI Innovator recently interviewed CEO Beerud Sheth to discuss how the business of conversational AI is developing in emerging markets – and also what’s ahead for the industry.
The AI Innovator: What problem is your company aiming to solve and how is your approach different from those of your competitors?
Beerud Sheth: Before talking about the problem, let me set some context. The internet emerged in an era of laptops and desktops, long before the smartphone was invented. As a result, users in developed countries are habituated to websites and browsers. However, if you look at emerging markets, messaging apps are the dominant digital interface, essentially replacing traditional websites and apps.
Users in these mobile-first environments find messaging apps more convenient and intuitive to use. You only need to look at WhatsApp and WeChat’s usage and open rates to understand how popular they are. And if customers are on messaging apps, can brands be behind?
For brands, it’s a huge opportunity to leverage the ubiquity and natural engagement of the conversational internet to build enduring relationships and drive business growth. When combined with AI, it’s set to cause a massive disruption in the existing ways of customer engagement.
Customers no longer respond to notifications and broadcasts, but have a strong preference for two-way, personalized interactions in real-time. This transformative paradigm necessitates a new integrated approach to support AI-driven journeys across acquisition, marketing, commerce, and support for brands.
However, existing enterprise software tools are optimized for web and apps, and they are not built to handle the richness of conversational channels and two-way interactivity. Furthermore, enabling these end-to-end conversational journeys requires a comprehensive set of new tools — from bots and generative AI for automation to interactive campaign builders, conversational commerce systems, click-to-chat advertising, and live agent capabilities. For most companies, these functionalities exist in fragmented silos across disjointed tools and teams.
Gupshup Conversation Cloud addresses this challenge by being a comprehensive SaaS platform that unifies all these capabilities, into a cohesive, AI-powered solution. Our platform offers businesses a holistic SaaS solution to design, execute, and optimize conversational experiences across the entire customer lifecycle, from initial acquisition campaigns and marketing journeys to commerce interactions and post-purchase support.
What trends in conversational AI are you seeing?
Conversational AI is experiencing transformative developments driven by advances in large language models and generative technologies. The most significant trend is the remarkable improvement in natural language understanding, with AI systems now capable of comprehending complex contexts and generating nuanced, human-like responses across multiple languages.
Multimodal capabilities have expanded dramatically, allowing AI to integrate voice, image, and text inputs, creating more dynamic and intelligent interactions. Personalization has become a key focus, with systems adapting communication styles to individual user preferences and delivering more contextually relevant experiences across various sectors.
Enterprise adoption is accelerating, with businesses integrating sophisticated AI assistants that can handle complex interactions and understand domain-specific terminology. Besides, the convergence of conversational AI with technologies like edge computing and Internet of Things suggests a future where AI becomes an integrated, intelligent interface transforming human-technology interactions across multiple domains.
Can you share some challenges you faced in developing conversational AI technology?
I wouldn’t say these are challenges but rather complexities that we’ve had to deal with. One such complexity has been language diversity. With a global user base, our AI models must understand and process numerous languages, accents, and dialects. This requires extensive training data that accurately represents the linguistic variations across different regions. We have invested in building robust datasets and utilizing advanced Natural Language Processing (NLP) techniques, such as transfer learning, to enhance our models’ ability to interpret diverse language inputs effectively.
Another challenge is user interaction complexity. Users often express themselves in unpredictable ways, using slang, idioms, or context-specific references that can confuse AI systems. To address this, we have developed sophisticated algorithms that focus on understanding context and intent behind user queries.
Additionally, ensuring data privacy and security is paramount for us as we work with financial institutions across the globe. While we collect and analyze vast amounts of conversational data to improve our AI models, we also adhere to strict regulations and ethical standards regarding user data protection, applicable in each of the geographies we are present in.
How do you ensure that your AI models are both accurate and efficient in understanding diverse languages, accents, and dialects?
We leverage NLP and Natural Language Understanding (NLU) technologies that are designed to interpret and respond to user inputs accurately. Our pre-trained NLU engine is capable of deciphering free text and natural speech queries across more than 30 languages.
Additionally, we focus on machine learning techniques that enable our models to learn from vast amounts of conversational data. By continuously training our models on diverse datasets that include various accents, dialects, and colloquial expressions, we improve their ability to recognize and interpret different forms of speech.
We also implement a robust feedback loop where user interactions are analyzed to identify areas for improvement. This involves recognizing new terms or phrases introduced by users and incorporating them into our models. Such adaptability ensures that our AI remains current with evolving language trends and user preferences, thereby enhancing its overall performance.
Finally, we prioritize user experience by designing our conversational AI solutions to retain context even during complex interactions. This capability allows our systems to manage mid-conversation subject switches smoothly, which is essential for maintaining engaging dialogues with users from varied backgrounds.
Can you share some specific use cases where your conversational AI made a significant impact?
While there are several, here are a few where we’ve been able to make significant impact on client’s revenue and sales.
CARS24, a prominent autotech company, has significantly enhanced its customer engagement and operational efficiency by implementing Gupshup’s conversational AI solutions. With the introduction of a Gen AI-powered commerce chatbot on WhatsApp, CARS24 has automated 80% of customer inquiries, drastically reducing the reliance on human agents. This shift has led to a remarkable 60% decrease in agent costs while achieving an industry-leading conversion rate of 8% for test drive bookings. The chatbot not only streamlines the car discovery process but also provides personalized recommendations, ensuring that customers receive immediate assistance and relevant information, ultimately improving their overall experience with the brand
Tonik Bank, a digital bank in the Philippines, has leveraged Gupshup’s conversational AI to enhance its customer service capabilities and streamline banking processes. By integrating a chatbot into their customer support system, Tonik Bank can now provide instant responses to customer inquiries regarding account management, transactions, and product offerings. This automation has significantly improved response times and customer satisfaction rates. The conversational AI solution allows Tonik Bank to handle a higher volume of inquiries without increasing operational costs, thereby optimizing their resources, and enabling staff to focus on more complex customer needs.
Reserva, an apparel e-tailer based in Mexico, has transformed its online shopping experience through Gupshup’s conversational AI solutions. By deploying a personal shopping bot on WhatsApp, Reserva simplifies the purchasing process for customers looking for clothing items. The bot offers tailored product recommendations based on individual preferences and assists shoppers in determining accurate sizing. This personalized approach not only enhances the shopping experience but also facilitates seamless checkouts on Reserva’s website. As a result, customers can make informed decisions quickly, leading to increased sales and improved customer loyalty.
NoBroker, a real estate platform in India, has utilized Gupshup’s conversational AI to streamline property searches and enhance user engagement. By integrating an AI-driven chatbot into their service offerings, NoBroker can provide potential renters and buyers with instant access to property listings and relevant information. This automation reduces the time users spend searching for properties while increasing the efficiency of the inquiry process. The conversational AI solution allows NoBroker to handle a larger volume of inquiries effectively, leading to higher customer satisfaction and improved conversion rates in property transactions.
How do you address concerns around bias and ethical challenges in conversational AI, including handling instances where your conversational AI might be asked inappropriate or sensitive questions?
Successful generative AI deployments at an enterprise scale are not limited to the quality of the output, but rather, output that’s reliable, in line with responsible AI principles, with controls in place to address issues of compliance, bias, hallucinations, and relevancy. Gupshup’s ACE LLM (fine-tuned foundational models’ layer) and Auto Bot Builder (tooling/chatbot layer) have built-in guardrails to help the bot stay on topic, provide contextual answers, comply with data residency/sovereignty/privacy laws, and safe deployment options. How we do this is by training the pre-trained models exclusively on a specific enterprise’s knowledge base, putting hallucination controls in place, strict boundaries on what sources the models are pulling responses from, among others.
We reduce the possibility of false or made-up responses, by restricting the source boundary to solely enterprise data sets. Enterprises also have the option to define limits on web crawling/URL searches (if trained on enterprise websites), which pages are included or excluded as part of the model training, defining confidence scores at the backend so the best responses is chosen by the model, with the option for the enterprise to add sources to responses to validate the model’s response picking technique.
What industries or sectors are you seeing the highest demand for conversational AI, and why?
Retail and BFSI (Banking, Financial Services, and Insurance) are two of the most in-demand sectors. In retail, conversational AI agents are being used all the way from marketing, product recommendations, re-engagement, commerce to support. Similarly, in BFSI, conversational AI is transforming customer interactions through applications like virtual banking assistants that handle account inquiries, fund transfers, and even fraud detection. By automating routine tasks, financial institutions are able to reduce wait times for customers and enhance service delivery, which is crucial in a competitive market where customer experience directly impacts loyalty.
Other than this, we are also seeing demand coming in from health care, education, automobiles, and travel and hospitality.
How do you see conversational AI evolving over the next five to 10 years?
Over the next five to 10 years, we’ll see a massive transformation across sectors, thanks to conversational AI. As organizations strive for more effective customer engagement, several key trends are likely to shape the landscape of conversational AI.
One of the most significant advancements will be in NLP. Future conversational AI systems will become increasingly adept at understanding the nuances of human language, including context, tone, and intent. The ability to process and respond to language with greater accuracy will facilitate a more natural interaction between users and AI, making it feel less like a machine and more like a human conversation partner.
Second would be hyper-personalization. By leveraging vast amounts of data, conversational AI will be able to tailor interactions based on individual user preferences and past behaviors. The future will also see a shift towards multimodal conversations, where users can interact with AI through various channels — text, voice, video, or even gestures.
Another exciting development is the emergence of emotionally aware AI. Future conversational agents will be equipped to detect and respond to human emotions, allowing them to adjust their responses based on the user’s emotional state. For example, if a user expresses frustration during a conversation, the AI could respond with empathy and provide solutions more thoughtfully.
Be First to Comment