The use of chatbots over the past 6 months has exploded with large corporations such as Google, Amazon and Microsoft making their chatbot technology platforms available for developers to build their own custom applications.

At Advantage Data we have been engaged to build a number of chatbots in health and finance and we expect this to be a core part of our business going forward.

How does a chatbot work?

A chatbot works by creating decision trees and at various branches different questions are asked to users. It’s a bit like the choose your own adventure novels that became popular in the 80s and 90s. Chatbots can be coded as purely rules-based engines with no machine learning or AI at all or you can use AI to inform the chatbot which branch of questioning to pursue next, based on the set of answers that the user has put in to date. We think one of the best applications of a chatbot is the medical diagnosis chatbot called buoyhealth. This system asks the same sort of questions that any GP would learning all the way and narrowing down on what condition you might have using AI.

As a starting point, the decision trees will need to be established. That’s often the hardest part if you do not have an existing dataset and are looking to collect one for the first time. Under such circumstances, it’s often the case that a rules-based implementation is setup first and experiments are made by changing the rules for a subset of users. This experimental approach allows diversification in the answers given and provides a framework for the chatbot to learn.

If you have a data set all the better you can start by seeding your bot with information that can be better than a rules based engine using AI.

What are the advantages of using a chatbot?

The most obvious advantage to using a chatbot is that they are great for data collection! You can ask more relevant and targeted questions to a user compared to cumbersome traditional surveys. Using the bouyhealth example, imagine how many questions would need to be asked to diagnose even the simplest health condition if the system could not learn from previous answers.

The other main advantage of using a chatbot is you can engage with your clients much better. The best chatbot applications give users relevant information all the way through. For example, bouyhealth givens you a list of possible diagnoses at the conclusion of the questionnaire, but that’s just one example. Financial chatbots might calculate the level of insurance that customers need or the types of investments that might be appropriate all tailored to the individual using AI. The best chatbots also have a feedback mechanism so users can report how relevant the information was that was returned so that the chatbot can learn from each individual interaction with a user. Setting up this feedback mechanism is an essential element to the success.

Should I use a chatbot?

In many cases investing in a chatbot is overkill, however, if you are collecting large amounts of valuable data such a service can improve efficiencies tremendously. For sales based businesses, chatbots can significantly help your conversion rate for your primary product or create cross marketing opportunities by understanding what products have the highest chance of success for a given user.

If your interested in finding out more about how chatbots work contact us and we will see if we can help.