Custom Software Development

A multichannel chatbot solution to the tedious process of requiring, submitting and receiving insurance quotes for multiple policy types (vehicle, building, home content and funeral insurance).

Building NIC

Problem:Applying for insurance quotes is a daunting and tedious process which can be intimidating for potential clients unfamiliar with the industry. The waiting period to receive a quote can be from twenty-four hours to a few days, leaving potential clients frustrated and less likely to sign up.

Solution:The Nedbank Insurance Chatbot (NIC) is an AI assistant that allows a user to capture information digitally via chat and receive a quote instantly once completed.

NIC banner

Project Overview

There was a clear need to streamline the process of applying for insurance quotes. With long waiting times and endless forms to fill in, potential clients were being lost.

To solve this problem, we created a platform where clients can easily access the information they need. Existing Nedbank policy holders are able to use the NIC platform to capture their information, receive instant quotes, submit claims, query claim-related FAQs, record complaints and contact live agents for updates or assistance. This eases the load on heavily burdened call centres and provides an easy-to-use interface for policy holders to interact with their insurer.

  • Building cover
  • Home contents cover
  • Funeral cover
  • Vehicle cover

Feature Highlights

The feature highlights pictured and discussed below include quote features, claims, and the service help desk.


As seen below, NIC currently supports four quote options. A user is also able to select ALL, which allows them to easily get quotes for all four options without having to re-enter required information common across multiple quotes.

Quote types

Shared quote features

Each of the options contain specific questions which capture the information required for an accurate quote.

Whenever captured information can be reused, such as client name and address, the assistant will prepopulate this information when completing the other quote types.

Date capturing via chatbot can be challenging because there are multiple format types and no calendar date pickers available. NIC is able to handle various date formats, allowing the user to enter dates in their preferred format.

When entering large numbers, NIC is able to interpret multiple number formats. For example, 350k = R350 000 and 1.5 mill = R1 500 000. This allows the user to easily complete the required fields for an accurate quote without having to worry about mistyping a zero.

Back and help navigational functionality is also offered throughout the user journey by typing “back” or “help” into NIC’s message window. Navigating back allows a user to select a different menu option or to simply edit a field that they previously answered. Typing “help” displays hint text for the current question NIC is asking the user. Live agent assistance is also offered via the “Contact Us” flow, or by directly asking NIC for a live agent.

Validation is shared across quote types as there is a variety of data types captured by NIC. The Atura Form Engine allows the current field being captured to be validated simultaneously so no invalid information is captured before submitting to the Nedbank API for the required quote.

Vehicle quote features

Vehicle quotes can currently be obtained for the following types:

  • Car
  • Motorbike
  • Trailer
  • Caravan

Once a user has captured their vehicle year model, NIC is able to populate a list of vehicle makes and descriptions that were manufactured for that year. This information is retrieved via the Nedbank Insurance API and is maintained regularly to keep an up-to-date vehicle database. The user can then choose a vehicle description from the drop-down list (see below).

Vehicle descriptions

Building and home contents quote features

When the user is capturing where the vehicle is stored at night, or simply entering their home address for a building insurance quote, a custom lookup webview is offered to ensure valid and accurate suburb information before submitting the quote request to the Nedbank API.

The webview is also able to calculate a partial match on the suburb. For example, “Pine” will yield matches of suburbs worded similarly, such as “Pinelands”. NIC will then provide the correct postcode for the suburb.

Suburb look up

Funeral Quote Features

The funeral quote allows clients to select a variety of cover options depending on their requirements. NIC supports quotes for multiple spouses, multiple dependants and extra funeral package options such as the Family Supporter Benefit and the Premium Waiver.

The following dependant types are supported:

  • Children
  • Parents
  • Grandparents
  • Siblings
  • Aunts/Uncles
  • Nephews/Nieces
  • Cousins
  • Domestic Workers
Completed Funeral Quote

Once the total policy cover, number of spouses, dependants and funeral package options have been captured, NIC is able to provide a quote, detailing the costs and user selections. If a user wants to change any of the previously entered information, they are able to go back and resubmit for an updated quote.

The Final Quote

Once the user has provided all the required information, NIC will return with various policy options offered by Nedbank Insurance for the user to choose from. A “Read More” button will display all benefit options for the selected policy and allows the user to email all captured details to themselves. Once they have selected a policy option, a live agent will contact them to finalise the process and ensure the client is happy.

Quote options
Quote options
Selected quote


NIC offers two options for claims: to chat to a live agent or to search the Nedbank Insurance FAQs regarding claim submissions. The claims FAQ functionality leverages Microsoft’s QnA Maker functionality to provide matches on keywords in the user’s text to map against claim related information in the knowledge database.


Service Desk

Users are able to ask for a live agent at any point in their journey. If the request is out of office hours, NIC prompts the user to leave a message and their contact details for the agent to contact them on.

Agents are able to see the full chat conversation with NIC which speeds up the help process. There are no queues when requesting a live agent; if an agent is unable to takeover immediately, NIC will provide the conversation to the live agent and ensure they get back to the user.

Live Agent Service Desk

Technology Used

The Nedbank Insurance Chatbot is an industry leader and first-of-its-kind, obtaining instant quotes within a single user chat session. It was critical for our team to create solutions that would support the functionality required for the best customer conversational experience.

Using current web technology trends and the Microsoft Bot Framework tool kit, the team was able to create a seamless conversation UX that allows a user to obtain quotes for insurance of multiple cover types, receive and review quotes, and make a selection for their desired business package, all within a single chat session.

The technology and custom solutions used to make it all possible are listed and explained below:

  • Microsoft Bot Framework
  • Atura Form Engine
  • LUIS Cognitive Services
  • Microsoft QnA Maker
  • Nedbank Insurance API integration
  • Webpack
  • SCSS
  • CosmosDB
  • Application Insights
  • Custom built webchat
  • Microsoft ServiceBus
  • Chronic DateTime Parser
  • Facebook Channel Support
  • ES6 JavaScript
  • Sendgrid
  • React

Using Microsoft Bot Framework as the core technology for this project was a simple choice due to the proven success and architecture integration it gave us on a previous bot project. At the time of development, however, Microsoft’s bot form engine FormFlow would not have been able to manage the various field types required by the NIC bot. As a result, we engineered our own form engine, the Atura form engine. This allows the bot to capture user input quickly, support over twenty data input types, validate data immediately, and allows the next required field to be rendered in chat.

We further chose to use Microsoft’s natural language processor LUIS to handle any free text that users would type into the bot. We chose LUIS mainly because of the close integration with the other Microsoft technologies at use. We also utilised Microsoft’s QnA Maker, which was implemented in the NIC project to automate Nedbank’s claims FAQs so that users could easily retrieve this information via the bot, without having to navigate the website.

Cosmos DB was selected to store important vehicle data for the quotes, suburb information, and live agent details. This allows the bot to quickly retrieve data needed to populate drop-down selections and be easily maintained and updated.

We used the Application Insights service for debugging and forming performance metrics across the application solution. With a number of middleware services, API integration, application handovers within the cloud, and messaging reliability; the team relies heavily on Application Insights to see behind the scenes.

Microsoft Service Bus was also used which, configured with a message topic, handles messaging routing between channels such as the web chat, application services hosted in the cloud, databases, and live agent handovers.

NIC requires multiple dates in different formats, such as driver’s license date, birthdate, residence occupation date, and vehicle model year. A dynamic solution was required to handle all this in conversational chat. With our custom code solution together with the ChronicDateTime Parser, it was possible to manage a variety of ways for the user to input this information.

The Sendgrid email service was integrated so that once a user has finished the quoting process, they can email the various options to themselves with a single click. This service is also used to notify support teams of any errors or drop-offs during a quote process so that the user will not have to restart the process if they are interrupted midway.

The service desk which live agents use to takeover chats when help or assistance is requested, was built with React Redux, allowing components to be easily updated and have their state modified while still being lightweight and fast.


A comprehensive overview of the architecture can be found on the Atura web site