By Harish Srigiriraju

Dedicated teams providing customer support can be expensive to maintain. Moreover, customer support teams need regular training to keep up with the product changes. Companies therefore, are implementing self-service options such as Chatbots to address customer issues.

A Chatbot is a feature that allows users to interact with the application through a messaging interface. Over the last few years, Chatbots have not only been used to provide customer support, but also to recommend products. Chatbots can provide an easy way to filter through a lot of noise on the application and quickly get to a specific offering the user is most interested in.

Chatbots can be designed through rule-based workflows that provide recommendations based on the users’ response to specific questions. Workflows can be tedious to develop and maintain if there are too many combinations.

Alternately, there are Artificial Intelligence (AI) models that can interpret the text from the user and provide necessary information. These AI Chatbots reduce the time taken for the user to quickly arrive at the information without having to go through a series of questions. More importantly, AI models, with more data for training, will get significantly better over time.

When using AI Chatbots, personalization is critical as the information provided needs to be relevant to each individual user. When a user is trying to trouble shoot or looking for specific products, general information is not helpful and can lead to frustration.

Therefore, it’s essential to monitor the performance of the Chatbot once implemented. Does the Chatbot satisfy the user by providing quick and relevant response? What metrics can be used to gauge the effectiveness of the Chatbot? These questions need to be answered before deploying any Chatbot.

Apart from monitoring metrics, user feedback is a critical step in refining AI models. One of the unique ways to personalize chatbots is to use In-App Ratings. In-App Ratings are collected when the user is using the application and is asked to provide a satisfaction score through a rating. These ratings are different from those collected on App Stores.

In-App Ratings can give a very valuable perspective on the user sentiment and can be collected at regular intervals. Once these ratings are collected, Chatbots can be personalized for satisfied and unsatisfied users.

Apart from providing usual customer support, Chatbots can provide additional offers such as discounts, extended trial period, and priority support to unsatisfied users to retain them. For satisfied users on the other hand, upgrades and cross selling can be the focus for Chatbots beyond customer support.

Without personalization, Chatbots can do more harm than good to user satisfaction. If personalization is not in place, companies should continue to rely on human intervention to resolve critical issues and can use Chatbots for non-critical ones only. Lastly, Chatbots can be more effective when In-App ratings are used as in input for the AI models. This unique approach will help companies retain the users longer and increase revenue in the long term.

Harish Srigiriraju is a principal engineer at Verizon. A graduate of Kellogg School of Management with an MBA.