Dive deep into CoreConverse – Max Life Insurance’s internal messaging bot


Max Life Insurance has an active base of 4 million customers with around 8 lakh of new policies sold each year, and the customer base is growing at a high rate of around 20% year-on-year. A large portion of our active customers write to us in email form on matters ranging from policy inquiries, policy clarifications, policy issuance status, shipping status, updated policy information and new service requests. On average, Max Life receives approximately 60,000 emails per month, which are answered by a dedicated email response team (mail office) that is part of customer service.

Due to the large number of email volumes, the current email response team requires a dedicated team of 50 to 75 people to review, queue and respond to emails on the base. manual review of each email. In addition, the response process itself becomes complex and requires queuing, scheduling, information validation and processing at the core workflow systems to ensure an accurate and rapid response on the job. basis of the complexity of the request raised by the client. This complete process with a dedicated email response team involves high operational costs. As a result, we are experiencing an increase in response TATs with increasing email volumes, which has a negative impact on customer satisfaction as measured by the NPS (Net Promoter Score).

The solution

In order to solve the above problem, we have developed an AI (Core-Converse) solution leveraging a built-in NLP engine that automatically understands the customer’s request by reading emails, identifies intents and provides initial resolution to the customer’s request. This artificial intelligence platform is end-to-end integrated with max Life’s core systems, such as the core policy app, website, and other customer-centric chatbots. This allows the platform to automatically identify the customer, chat history, validate customer information, extract relevant information required for customer’s query and send it back to customer as a response by e -mail. Additionally, in addition to responding to customer queries, the platform also manages the entire email workflow by scheduling and queuing responses to ensure the minimum possible turnaround time. to resolve these queries.

Currently, the Core-Converse AI platform automatically processes 50% of customer email queries with an accurate 90% response rate (10% of responses are reviewed manually). different kinds of customer queries for which customers write to us. For the remaining 25% of queries, which are quite complex in nature, such as those related to policy buyback or a combination of multiple queries that a client raised in a single email, the platform creates a response that is currently being reviewed manually, then sent to the client. With continuous improvements, we aim to make the NLP engine more accurate to handle over 90% of customer queries automatically through the platform.

This has allowed us to reduce the cost of the courier office by almost 30%, reduce the average TAT by 90%, and improve our customers’ NPS by around 1%.

How it works

The diagram below represents the overview of the process managed by the Core-converse platform.

C:  Users  vshom4069  Desktop  email bot.png

The process begins with receiving the email sent by the customer. After the email is received, the validation engine program identifies and matches the email id of the customer by looking for it in the registered customer information. If the email id is recognized as one of the existing customers, the email goes through the main NLP engine, which identifies the intent of the email. (what information is the customer looking for).

In order to resolve the customer’s query, the engine extracts relevant information from the text of the email, such as policy number, customer number, and other required information, such as if a customer is looking for a receipt from policy renewal payment as intent. These inputs are then provided to the main Max Life applications to get the information needed to resolve the customer’s query through various built-in API calls. Once the response information is received, the platform integrates it into a personalized response email template and sends it back to the customer. The main converse platform automatically performs these steps by managing the end-to-end email receive-reply process.

The full platform is deployed in the cloud leveraging both the AI ​​and the AWS and Google cloud component deployment tool stack. The following figure shows the technical architecture of the solution.

C:  Users  vshom4069  AppData  Local  Microsoft  Windows  INetCache  Content.Outlook  CEY4UAJI  Picture2.png

Architecturally, the whole platform is divided into two main parts: the messaging workflow system and the Core AI engine.

Email Workflow System –

The main purpose of the messaging workflow system is to identify and validate the sender by retrieving information from various source systems. Using inputs from the main AI engine, it automates the entire messaging process. After the email arrives, the main NLP engine extracts information such as email id, domain name, policy number, and intent from the email.

See also

The workflow system is connected to MLI source systems via APIs to validate identity and retrieve all execution-related data. With input from the main NLP engine, it validated customer identification through the policy number and registered email id, while retrieving the required customer information / data and supplies to the main NLP engine to create the reply.

During the validation process, the workflow component also identifies if the customer has sent repeated emails for the same issue, if the email is tagged with other specific email ids indicating escalation and whether the sender needs generic information, such as office address, office contact number, etc. This complex information allows for proper planning and prioritization of responses in the email response queue.

Basic AI Engine (NLP) –

The core component of the platform is the NLP-driven AI engine. This mainly involves understanding the content of the email through NLU and extracting client intents and other relevant information. The engine uses Dialogflow as the primary NLP engine. It is surrounded by a custom layer of preprocessing engine deployed on AWS Lambda, which functions as a tightly knit unit to understand customer intents from the email body and responds to the CoreConverse workflow with intents. identified. In order to overcome the limitations of the number of characters that can be handled by the dialog flow, there is a custom trained template to split large emails into smaller snippets so that all emails can be processed. Before the main engine intends to identify, this custom trained model breaks up the email into pop-up snippets and sends each of those snippets to the Core NLP engine, which responds with the customer’s intents, which are again consolidated and cleaned up. to identify any multiple intent and remove any repetitive intent. During this extraction process, the model performs subject modeling by identifying part of speech, lemmatization, spell checking, word similarity, and frequency analysis.

The platform uses a full technology stack. With the Google Dialog Flow NLP engine, the custom model is built in python. The platform’s front-end was developed on ReactJS, and a range of managers, reporting services, workflow – task – decision, pooling, and staging services were used for deployment to the AWS cloud. Additionally, Netflix persistence has been used to keep the workflow stakes.

Subscribe to our newsletter

Receive the latest updates and relevant offers by sharing your email.

Join our Telegram Group. Be part of an engaging community

Leave A Reply