Don’t Underestimate the Power of an AI Chatbot (Part 1)

Authored By Utpal Chakraborty, Head of Artificial Intelligence at YES BANK, AI Researcher, Speaker, Agile & Lean Practitioner

The experience interacting with a Chatbot at many occasions for many of us probably was not that convincing despite it is being marketed with catchy tags like powered by AI, Smart & Intelligent, Self-Learning, Self-Enriching etc. And many people have already drawn their conclusions out of frustrations that Chatbots are not at all intelligent; and behind the scene a bunch of hard coded logics are working to facilitate the conversation- Which was probably partially true for some cases few years back. At large the story today is a bit different and we need to understand both the sides of the story to draw a fair conclusion. We will try to discuss different aspects of it one by one.

But despite all these, Chatbots are still “Hot” and their adoption across all industries has increased many folds in past couple of years. And as per the experts & analysts, the craze for Chatbots and virtual assistants is going to continue in future and they are going to be the primary mode of interaction and collaboration with the internal as well as external entities for any business.

Broadly there are two reasons for the past customer complaints concerning the usability of the Chatbots and their maturity.

Firstly, I think we must admit that technologically we are not yet at that level of maturity to provide a complete human like conversational experience with a Chatbot; although many of us have already claimed so desperately. We are certainly progressing towards achieving such state in a very fast pace but right now probably we are not there. As a Chatbot platform provider, we should have set the expectations right in front of the customer rather than over-hyping it for no reason. We still need few more years to arrive at that stage or anything near to it wherein our Chatbots will be capable of providing conversational experience as good as a well-trained customer service agent. So the things are getting better, getting matured and stable every day incrementally and we are not that far from what we have been expecting our Chatbots and virtual assistants to be for so many years. Google, Amazon and many others are investing a lot in this area alone as they have realized that this area is hot and it’s going to be hotter in coming years.

Secondly, although many of the Chatbot providers in the past had claimed that their Chatbots are powered by a strong NLP Engine (Natural Language Processing Engine); but in reality either there were no actual NLP Engines at all in many of the products. Or even though they had one, probably it was not up to the standard for handling customer queries the way it was expected to do so. The core of a Chatbot is the NLP Engine, and the technological maturity of the NLP engine within it essentially decides the efficiency of the Chatbot.

The other side of the story is, every organization wants it’s interaction with customers or for that matter interaction with any other internal and external entities to be in a controlled manner, something referred to as “Controlled Conversation”. You can correlate it with any human customer service agent; he or she has hardly any liberty to speak beyond few hundred sentences. So, we need to also understand the fact that there’s going to be some boundary and context beyond which Chatbots are not expected to be autonomous. For example, a Chatbot powered with a deep learning model, trained with few years of conversational data from some social media platform like Facebook or Twitter; cannot be deployed in a business environment because it would probably try to cross the boundary of that “Controlled Conversation” because it has been trained with such unrestricted conversations which normally happens over social media. Rather a Chatbot in such delicate environment is expected to gracefully take the conversation to a logical end without crossing the predefined boundary – “The Lakshmana Rekha”. Here a fully autonomous and super intelligent human like Chatbot is probably of not much use at least with the business context. That’s where some amount of logic and restrictions comes into picture to make a tradeoff between what Chatbot can do and what is actually expected out of it in that particular environment. But the Chatbots and Virtual Assistances of today are far better than few years back and I am sure those of you who are keeping a close eye on the advancements of NLP engines is going to agree with me. Believe it or not, recently I was interacting with a Chatbot and it was amazingly intelligent, I could recognize it as a Chatbot and not a human agent almost after 30 minutes of conversation.

Unfortunately there has not been much work done on industry benchmarking and grading of the Chatbots and their features till date. In this article I will be trying to touch upon few areas and some of the critical features which almost every industry and customers are expecting from a Chatbot today.

An Intelligent Chatbot features can be broadly grouped under following heads and we will be discussing each of those in a bit more details in this series.

1. Conversational Maturity
2. Handling Contact Centre Functions
3. Advanced Linguistic Features
4. Personalization Features
5. Smart Automation Features
6. Smart Recognition Features
7. Co-Browsing Features
8. Emotionally Intelligent Features
9. Omni-Capability
10. Accessibility Features
11. Integration Features
12. Analytical Features
13. Advanced Collaboration Features with AR/VR
14. Hierarchical Features

1) Conversational Maturity:

First of all a Chatbot is expected to be able to take the conversation with a user to a logical end. And if in case not by itself, it should be able to transfer the conversation to a live human agent seamlessly so that the continuity of the conversational experience with the user is maintained. It is also expected to create a service request in case where user request needs more investigation and expertise to resolve the same. An intelligent Chatbot is expected to be able to take on personality traits of users, it is interacting with and customize the response on the fly to the characteristics of the person it is talking to.

Another important feature is it should be able to take feedback from the user on the conversation instantly, whether the user query has been resolved or not before the completion of the conversation.

Other areas of conversational maturity counts where Bot should understand when the user wants to “Restart the Conversation” into a completely different conversational context or track. Similarly it should be able to prompt the user before the session expires to increase interactivity.

But the one most important feature expected from any intelligent Chatbot is the feature of continuous learning and self-enriching, technically which is a bit tricky. But features like Feedback Learning, maintaining Domain Vocabulary and periodic training with newer vocabularies can surely help bringing some level of continuous learning into your Chatbot.

2) Handling Contact Centre Functions:

One of the most convincing role the Chatbots are playing today is taking over the contact center functions. While it’s been quite successful in terms of cost optimization and providing 24×7 support to the customers across geographies but on reality we are still far from that ideal state of relying completely on Chatbots to handle all customer interactions and transactions. Here “Transfer to Live Agent” feature is very essential. Chatbot should be able to seamlessly transfer the chat to Live Agent with complete history of the conversation wherein the live agent should be able to chat with the user in the same chat interface as the Bot. Also, features like “Click to Call” should be available to the user in the Chat interface itself in the event if no live agent is available to take the chat-control in case of out of support hours. An outbound call request or callback workflow should be triggered as per the suitable time opted by the user for the callback.