In late 2015, Yellow Messenger developed a location-based messaging platform, enabling users to find nearby businesses and engage with them in real time. For instance, a user in Koramangala, Bengaluru, could locate all companies in the area via the app and hold a live chat, with responses coming from the businesses or the Yellow Messenger team.
This added value to customer communication at a time when few CPaaS (communication platform as a service) companies offered conversational AI to enhance customer support. But Yellow Messenger’s founders – Rashid Khan, Jaya Kishore Reddy Gollareddy and Raghu Ravinutala – soon realised that a traditional chatbot/text messaging model at the front of the contact centre lacked innovation and scalability.
“We noticed that users repeatedly asked similar questions: When are you open? Do you have this item in stock? How can I return a product? By 2016, we had 50K active users. However, our work was manual – we gave businesses an agent app or kept responding ourselves,” observed Khan.
In contrast, fully integrated artificial intelligence (AI) can transform customer journeys through 24×7 multilingual support. This helps automate a significant portion of the enterprise workflow, resulting in greater customer satisfaction, more lead generation and increased productivity by freeing up customer agents’ time. Better still, AI chatbots can analyse data to create personalised responses, focussing on context and intent, which is a huge value addition. So, the trio joined Microsoft’s accelerator programme in the same year.
Aware of the business potential, the team at Microsoft encouraged the founders to change their consumer-facing app and eventually build a customer service automation platform for enterprises. Its initial clientele included prominent names like Indian Oil and Asian Paints.
Embedding AI for a seamless customer experience was a challenging task, though. According to Khan, AI-ML-based automation was pretty rudimentary in the beginning. It was a rule-based system that could answer five, 10, or 20 queries, but that was still a lot of value addition for many businesses.
However, there was a need for evolution, and the founders opted for a platform-based approach in 2017 so that companies could build and manage bots and send notifications. A human agent can also address queries if a bot response is escalated.
AI adoption became easier when Google made several open-source natural language processing (NLP) models during 2017-18.
“For instance, it launched BERT, a machine learning framework for NLP, and we used that initially,” said Khan. “But gradually, we moved away from Google to a lot more ML-based intent processing and action-driving models [to enrich our offerings]. We finally forayed into generative AI (genAI) in 2021 and rebranded as Yellow.ai.”
The newly launched brand represented the company’s evolution to delivering a single platform to automate customer service throughout the customer lifecycle, with AI at its core.
Headquartered in San Mateo, California, Yellow.ai today serves 1,100+ enterprises in 85 countries, including Pelago (part of Singapore Airlines), Waste Connections, Sony, Domino’s, Hyundai, Volkswagen, Decathlon, Randstad and the Lulu Group International, among others. It offers products across 35 channels (text and voice platforms like mobile, social media, messaging, web, voicebots and more) and supports 135 languages. It also claims to automate 16 Bn+ platform conversations annually.
Yellow.ai claims to have achieved 5x rise in global revenue in the past three years and built a team of 650+. It claimed $30-40 Mn in annual recurring revenue (ARR) in FY24 (February 1, 2023-January 31, 2024), with India accounting for 30-40% of its business. According to Khan, the platform deployed 120+ genAI bots in FY24.
Other key markets include North America, APAC and the Middle East. It is also aggressively expanding its business in the UK, the EU and Latin America. The platform competes with the likes of Gupshup, Reliance-owned Haptik and Verloop and eyes a revenue growth rate of 80-100% in FY25.
What Has Pushed Yellow.ai Towards GenAI Adoption
As Khan explains, customer satisfaction matters most in reducing churn and increasing conversions, thus leading to higher CLV/CLTV (customer lifetime value). Moreover, predictive analysis based on large datasets can enhance service quality and enable proactive measures to drive sales.
Additionally, an AI-driven customer service platform integrated with the enterprise workflow accelerates the automation level expected from Industry 4.0 companies.
Yellow.ai keeps a tab on these innovations and technologies to help enterprises in the middle of an AI-powered race, and genAI turns out to be one of the best-performing, envelope-pushing models in the foreseeable future.
The earlier chatbots built by Yellow.ai required a lot of training and could only respond to specific queries. The startup also realised that most enterprises did not have suitable datasets or they did not want to put much effort into training each bot.
Besides, in the last year and a half, technologies have gathered pace and evolved swiftly, and the founders realised that their earlier products became obsolete compared to genAI advantages.
“We then started working on internal problem statements like: Can we start generating training datasets so that our enterprise customers don’t have to spend time on it? That was our first foray into generative AI,” said Khan.
How Yellow.ai Is Automating Customer Service Suite To Empower Enterprises
To explain how customer support differs in the AI era, the founder came up with a simple use case. Suppose a customer wants to cancel her hotel booking. Previously, she would have to call the contact centre, wait in the queue (probably for a few minutes) and then speak to an agent for a resolution for the next five to 10 minutes.
She can submit her queries to conventional AI chatbots but will receive a response only if the bot has been trained accordingly. It is usually 30-40% effective.
With genAI, companies need not train every bot for every use case. This is because generative AI uses large language models, or LLMs, built on a neural network that helps generate near-human contextual responses. Therefore, genAI boasts intelligent bots that respond authentically and dynamically, driving conversations without the need for specific queries inputted in the system. Instead, one can just focus on the business goal it is trying to achieve through the conversation.
In essence, with generative AI, one could throw a curveball at an AI-based bot, and it would handle it 80-90% better than a conventional AI bot.
“It means we just come and say: Hey, the goal of this particular workflow is to be able to cancel a hotel booking by collecting this information and integrating it with these systems,” said Khan. “We also started thinking a lot about how we should use genAI in the context of solving customer support problems for large enterprises, reducing their costs and bringing in more automation.”
Yellow.ai’s Product Suite And Its Industry Impact
Yellow.ai is now delivering AI-based omnichannel solutions for a seamless customer experience. As Khan explains, an omnichannel approach goes beyond isolated channel experiences and focusses on a holistic customer journey.
For example, if customers initiate contact through social media and then transition to the company website or switch to an app or a phone call, they can expect a consistent information flow and personalised experience throughout their journey. Such measures are only possible with an AI-first omnichannel customer service, which enables businesses to streamline processes, capture and synthesise customer data across channels and deliver timely and relevant responses.
The full-stack unified platform features a host of gen AI-enabled solutions such as a help-desk ticketing system and support for CRM, chatbots and voicebots.
Other offerings include:
- Auto summarisation: Customer support agents can quickly access the entire context and current status of a query through a summary, and customers need not explain their issues multiple times.
- Auto tagging: Due to the high volume of incoming customer requests, it can take time to determine the nature of the conversation (information gathering, issue resolution and more). To overcome this challenge, the platform automatically identifies and labels the trend or pattern behind the chat using keywords called tags.
- Auto sentiment analysis: With the help of a summarised ticket, this feature can highlight critical points of a past conversation and attribute a positive or negative sentiment to the outcome, enabling the agent to understand the customer’s state of mind and respond accordingly. In fact, it goes one step further and provides response and tone suggestions.
- Knowledge management: This feature enables enterprises to effortlessly optimise and convert their respective knowledge base into an intelligent bot without any additional effort.
Khan claims each customer support interaction via chat costs $8-12, while Yellow.ai’s AI-powered solutions cost less than $1. Again, a voice interaction costs between $15 and $25 per call, while the cost at Yellow.ai is limited to $1.5-2.
Overall, it claims to have achieved a 40% rise in customer satisfaction score (CSAT), a 60% reduction in operational costs and a 50% increase in agent productivity. Although calls are routed to human agents depending on complexities, the startup claims up to 90% of calls are addressed by genAI bots.
Yellow.ai’s Product Roadmap And The Journey Ahead
Going forward, the enterprise-focussed genAI platform aims to enhance the intelligence and capabilities of Yellow-G. It aims to:
- Launch voice AI with a built-in voice LLM; this will help automate contact centres from 10% to north of 80%.
- Build more customer support-specific LLMs that can handle support tickets by understanding and orchestrating workflows, eventually leading to issue resolution without transferring queries to human agents
- Develop Yellow-G, a generic LLM platform, to help enterprises build their genAI models for specific tasks
Although most genAI bots are limited to text-only conversations, incorporating visual elements can significantly improve user experience. The Yellow.ai team is now exploring multimedia support for LLMs to create design-centric and experience-focussed bots to delight users.
“To improve customer experience, we are introducing auto-generated quick replies, images and videos from knowledge-based content. We will also use various forms of visual communications like carousels and GIFs,” said Khan.
Also, Yellow.ai is focussing on voice quality, aiming to do away with the robotic tone often associated with text-to-speech engines. The goal is to achieve more natural, human-like voice outputs and support enhancements in the next few months.
“Other focus areas include ongoing R&D on LLMs and optimising their speed and efficiency with reduced training requirements. This represents a longer-term focus where Yellow.ai is dedicatedly pushing technological boundaries,” added Khan.
Generative AI is no longer the next frontier but a fast-emerging trend that has overwhelmed every industry and business due to its speed and reach. The conversational AI market, globally embraced by most enterprises for efficiency and cost-effectiveness, is also growing in sync, from $10.7 Bn in 2023 to an estimated $29.8 Bn by 2028, at a CAGR of 22.6%. Although the India numbers are not out yet, media reports predict a growth rate of 20-25%, which means global players like Yellow.ai will focus even more on this market.
The only glitch: Will enterprises part with their sensitive/proprietary data so that third parties can build business-specific genAI models for them? Or will they explore open AI and generic LLMs to make customised solutions? Regardless of the outcome, GenAI platforms like Yellow.ai and its peers with an innovative bend will emerge winners.
[Edited by Sanghamitra Mandal]
The post How Yellow.ai Is Crafting GenAI-Powered Customer Support To Strengthen Enterprise Growth appeared first on Inc42 Media.