Revolutionizing Customer Service: Airtel’s AI-Powered Chatbot
Business Context
At Airtel, we encountered a substantial challenge in catering to our vast customer base. On a daily basis, our contact centre received an immense number of calls, emails, and messages from social media platforms. A lot of these queries were from app users, and our digital channels did not have the capabilities to adequately service their needs.
To assist with this mammoth task, we introduced — a conversational AI chatbot under Help and Support section. This had an incredible outcome, with thousands of daily users taking advantage of our virtual customer service. Not only did it reduce the number of people resorting to calling us, but it also gave users the chance to resolve problems independently using our FAQs and context-specific support options.
Problems to Solve
Airtel has a mission to ensure their customers receive the best service possible in the simplest way possible. This involves offering digital help and support that is tailored to the customers’ needs. Customers will be able to quickly access the information they need, take the necessary actions, and receive helpful support.
1. To Unify Support Workflows in One Centralised Hub
Airtel and the entities associated will have one centralised hub from which they can execute support workflows, share essential details, and determine the effectiveness of the processes and flows.
The experience team will have one channel from which they can provide generic information about all services, provide contextually relevant help for non-Airtel customers, and gain insight regarding process and flow effectiveness.
2. To offer comprehensive customer service platform for queries, complaints, and more
Through this platform, Airtel customers can easily access support related to telco services, broadband and DTH services, Airtel Payment Bank and Wynk accounts. Whether customers need to ask a question, find information, request an action to be taken, submit a complaint, or follow up on their issue until resolution, they can find assistance quickly and easily.
A platform that promises to provide Airtel customers with the assistance they need in a convenient and accessible fashion.
Solution
At Airtel, we have created a conversational AI-based chatbot platform that enables users to take advantage of the Airtel Thanks app, WhatsApp and other channels all in one place. The platform features a trainer module that uses a multilingual language model to detect and map intents, as well as fulfilment flows. This model is trained on a diverse set of multi-lingual data.
Our conversational AI framework simplifies the process of creating complex conversations. Our model comprehends multiple languages, perfect for widespread implementation. Furthermore, it uses advanced algorithms to carry out natural language processing (NLP) to guarantee a seamless and tailored communication. As a result, users are presented with an effortless and personalised conversation experience.
On the chatbot platform, you can take advantage of the many features to enable a smooth conversational flow. You can create forms and workflows, define the slots and data types for each one, and even create branching stories based on different conditions.
Additionally, you can create rules to map to the actions defined on your action server and use our settings panel to enable the NLU and other relevant endpoints. Our CRUD admin interface allows you to inspect the allowed intents and mapped action scenarios, while our vernacular admin panel lets you easily inspect and manage ingested translations.
Let’s gain an understanding of how the NLU Engine and Workflow Builder on our platform operates.
NLU Engine
As a user engages with the chatbot, the conversation flows through a series of steps that leverage the power of machine learning and natural language processing.
Firstly, the chatbot obtains the user’s message and applies techniques such as feature extraction to identify the most pertinent aspects of the text. These aspects are then provided to a trained language model, which leverages its familiarity with language structures and meanings to anticipate the user’s intent. This prediction is made using in-house models and a natural language understanding API.
Once the chatbot has determined the user’s intent, it uses dialogue state management to select the most appropriate response. This response is carefully crafted to address the user’s needs and continue the conversation in a meaningful way. Finally, the chatbot sends the processed response back to the user, completing the cycle of communication and allowing for a seamless and engaging conversation.
Workflow Builder
Our workflow builder allows you to easily hook third party APIs and inspect the responses right on the user interface. And with our utter actions and responses configuration and training support, you can fine-tune your stories and mappings for even more personalized and effective conversations.
Let’s look at the workflow where we make use of our conversation builder to create a WhatsApp channel to service prepaid lead creation users via integration with Airtel IQ.
The Bot Platform is fully equipped to help organizations streamline their customer experience. Our AI based conversational system offers single-point access to customer journeys, consistent messaging across all channels, and an end-to-end view of customer engagement.
And this is how our chatbot platform appears in its simplest form.
This comprehensive suite of features allows developers to design, create and debug conversational flows in one place.
Conclusion
Using our advanced conversational Artificial Intelligence (AI) bot, we are achieving faster and more efficient customer service, while simultaneously reducing operational costs. In addition, this technology enables us to reach a larger base of customers.
The team is actively researching and implementing AI and ML algorithms to enhance the customer experience. We believe that AI is an important part of the future and we are excited to bring it to our users.
Acknowledgements:
I would like to express my gratitude to my colleagues Shekhar Govind, Bharat Shetty Barkur, Suman Malhotra, Harsh Jhawar, Poornima Kapoor, and Kishalay Bose for their help in making this article possible. A special shout-out goes to our mentor, Rohit Jain, for his invaluable guidance.
