AI for CRM Business Series : 4

Growing consumer base is the key to successfully take off your business and in this more than ever connected world, the word travels out faster than ever – good or bad. Thus, customer support, especially pre sales and after sales service, is crucial to sustain and grow the customer base with excellent business reviews. In this day and age, apart from products and services, customer relations position itself at the center stage which can completely take your business game to new levels.

The key to successfully grow and sustain the business is to spread a good word and especially leverage technology in the process of bringing satisfaction to the customers. The best way to achieve customer satisfaction is to always listen and swiftly act to resolve issues. Today, with technology empowering businesses and individuals, we are positioned closer to our consumers, however with the limitations of resources, especially support teams we must prioritize customer issues and hear out all customers each time and bring out resolutions quicker. Well, to cater to this arena of listening, resolving and prioritizing customer cases, with limited resources in place, it is essential that we smartly use these support resources to bring maximum customer satisfaction and also use them in in an ecosystem where technology empowers the customer support teams.

This brings us to the interesting topic of ChatBots that has been empowering Customer Support. With AI making progress in all arenas, ChatBots also seem to be smitten by Artificial Intelligence, especially the sub-domain of Natural Language Processing capable of understanding and interpreting human language. The sub-domain of Natural Language Generation helps machine generate relevant replies and answers based on its training (essentially lots of Data). The two subfields within AI seem to quietly bring change and might forever transform the CRM space – especially the Customer support for better. Let’s understand this to lock gears with the changing technology and times for redefining the Chatbot space.

What is a ChatBot?

ChatBot is powered by predefined business rules, scripts and artificial intelligence through a conversational interface [1]. Within the last 10 years, the ChatBots have transformed from a rather automated tool working in a business ecosystem with predefined business rules to a more intelligent support partner, capable of making smart decisions to solve business problems, directing users to suitable resources based on historical data and have gathered enough intelligence to understand human language and respond with suitable answers and improvements continue to be made in these directions.

More recently, the Salesforce Einstein ChatBot has helped solve on an average about top 3 customer inquiries (common product or service issues) in the chat channel. Such inquiries make about a third of the total overall chat volume. This has helped reduce average call volume for individual customer representatives. Thus, each can prioritize their time to solve rare, tricky issues, have more time to bring service quality with fewer issues and perform value added activities. Lesser agent handle time also leads to smaller customer waits for better experience and happier customers! [1]

Understanding Your Business Transactions

However smart the ChatBot is, it is significant that it is taught and especially trained to perform the necessary actions. Thus it is very essential it is trained with the correct kind of inputs (customer oriented and business specific), and given appropriate time to undergo training over historical data sets before being put to deployment. This can be analogously compared to the experience and education of a newborn to an intelligent adult.

The ChatBot must be trained based on diverse inputs from various customer facing business functions. In order for the ChatBot to sort out the order of things for business, here’s some homework. Analyze frequent issues faced by your chat agents or support teams and how simple is it for your bot to handle them based on the complexity. These “actions” are termed as “transactions” that will serve to act as the ChatBot’s building blocks. Over a period of time the inputs must be updated to challenge the Bot more for better performance.

These “transactions” shall receive one or more than one input and shall also deliver output(s). These transactions essentially become dialogues for the customer and can have multiple exchanges based on the Menu (decision tree like algorithm) governing the responses.

The Einstein Bot builder offers configurable point-and-click setup to allow building dialogs in the Bot. It supports different types of tasks that can be configured into dialogs. The Process Rules can help configure the flow of dialogs based on logic. Certain processes can also be automated based on the Run Scripts to store answers in Salesforce or external systems.

Customer Intent

A Natural Language Processing technology based bot is also possible but requires extensive training before satisfactorily being able to function for a business use case. Understanding customer intent can be a tricky thing, because a single intent or question might be asked in multiple ways. Thus the training of Einstein Bot in such cases requires multiple possible inputs which one may not get right in single attempt and thus requires fine tuning with multiple “training” iterations. In each of the dialogs configured, a separate Intent tab gives you the place to enter these commonly used phrases by customers. The goal is to however make the bot understand similar questions asked in multiple forms. This brings you to your second homework.

To begin with, it is advisable to use the Menu feature with click-off dialogs to start with and the Intent feature helps to build the Bot strong capable of theoretically answering anything based on understanding the language and customer intent over time. For unexplored phrases, words and questions the case is elevated to the Sales agent. Later on the Menu can be removed, once the Einstein Bot is completely trained to deliver strong performance.

ChatBot Deployment

The ChatBot shall be deployed for the larger audience after it gets a perfect score in the User acceptance test. Chat buttons or proactive invites typically reside on the website or mobile app. Einstein Bots can be configured on top of the live agent. Thus, when a customer clicks  the button, they initiate a conversation with the ChatBot which at any stage can be manually “Transferred to an agent” or gets elevated to the agent with right skill set, in cases the bot is unable to answer at any stage. The chat transcript with the bot can be reviewed by the right agent, thus the conversation doesn’t end up being redundant and is very helpful to the customer.

ChatBot happens to be an interesting piece of technology being transformed by AI. However, it needs to be seen if it shall become a necessity for business. Early technology adoption has by large been fruitful. How shall it transform the industry or change the way customer support functions needs to be waited for and seen. But do you feel the impacts of AI transforming the ChatBots and the CRM space ?

Feel free to share your views and comments on the topic. Also, keep looking for more on AI in the CRM space in this section.

References:

  1. https://www.salesforce.com/blog/2017/11/the-magic-of-einstein-bot