Believe me or not, you are perhaps going to get into the era of AI (aka artificial intelligence). It is not a digital transformation you have heard over the Internet. It’s personally an intelligence transformation when your computer *would* understand you, as known as computer vision. One of the computer visions that folks pay pretty much attention to the development is conversational agent. Unless you like to call chatbot. Last weekend I sat down with a friend of mine to muse whether conversational bot gives you a good return on investment or not.

What is a conversational bot?

Firstly, let’s get just a little bit understanding of what a conversational bot in the era of AI is. It’s a virtual agent running on your computer or via web browser that can understand your question. For example, every time you need to book a flight ticket, you have to go to the intranet portal or send an email to an administrator to seek her support. This process would take time to wait for the response. With conversational bot, it’s like a virtual administrator who can understand your request via instant messaging tool (e.g, Skype), and rapidly help you. Instead of filling in a formal electronic form, just go ahead chatting something like “I need to book a flight ticket to Singapore next week from 20/01/2017 to 23/01/2017”. If that virtual administrator understands this sentence, it will then ideally return to you a new message like “Would you prefer any airline?” or perhaps “Would you prefer any hotel?”. Moreover, it could return to you 3-4 accommodation options. Another example is customer service inquiry. Imagine if you have 300 customers asking your set of products every hour, how many customer service staffs would you prepare to answer those questions? This is a very common issue today and conversational bot can help. With such a virtual staff, you do not have to hire many people who just sit down at computer chatting with customers.

I’m not going to dig into components in a conversational bot. In the simplest explanation by technical language, a conversational bot has NLP (aka natural language processing) algorithm to assist software or back-end code understand your informal questions. When it understands you, it needs to vet through historical data before properly answering you. If it does not understand you, it will have to learn with ML (aka machine learning) algorithm until it understands you. The term “Understand” means by accuracy. It’s like a conversation between you and a kid. If he understands your age inquiry, he will answer you. That is what I myself understand about conversational bot today.

Good return or bad return on investment

The big problem I’ve seen today is the perception of rolling out conversational bot in a production environment. Acknowledge the fact that you can very quickly build a conversational bot within 5-10 minutes with given AI services from giants such as Facebook Messenger, Microsoft Bot Framework, IBM Watson or so on. But remember that is just a POC (proof-of-concept) to showcase to your customer or your boss. Think about a deployment of conversational bot for technical customer service for Internet provider. If the bot does not understand your customer, would your customer be still interested in your service? How does the customer churn get increased? Saying my internet connection is slow and I ask a conversational bot to chat with virtual technical engineer. Are you sure your conversational bot understands my question, and if it understands, will it properly give me a guide to support me? If the virtual technical engineer misunderstands then will you be still happy? Compare with an actual well-trained technical engineer who understands and well communicates with a customer.

Back to my conversation with the friend last weekend, we talked about the ROI of intelligent flight booking system which allows you to chat directly to a back-end conversational bot to book a flight. The bot asks you to choose your journey, date and time including preferred hotel at the destination. Things sound interesting. We had a serious question: Does such an app make me be satisfied? What is the actual problem with current booking system on the browser that we normally do in life? Is there a problem with calling to an agency to book a flight when I just need to say my journey, date and time and wait for the support? This came to us a conclusion that this kind of intelligent flight booking system using conversational AI is not more helpful in my real life. It may let me down if the app is stuck on understanding me and then lets me wait for an unacceptable time, when I need to urgently book a ticket.

Are you happy to spend more time to chat with this bot?

Skyscanner offers a bot for flight search: Chat your way to your next trip: Skyscanner launches Facebook Messenger bot for travel search

I like the idea of conversational bot in order to increase customer experience, but almost this is likely a conceptualization. And organizations are seriously careful in the production deployment because this badly impacts on the customer experience tendentiously. If you make just a little mistake, satisfying your customer again is likely impossible. This is the rule of business. I would personally think conversational bot is only helpful for information inquiry today, or just very basic questions (e.g. product guide, user guide…) rather than a more complicated system replacing human.

In a nutshell, ask yourself the following questions, before rolling out your conversational bot in the production environment:

  • Can your conversational bot really replace human work right now?
  • What kind of information does your conversational bot answer?
  • If a conversational bot misunderstands, how does this impact on your SLA (Service Level Availability)?
  • How long does it take for an answer, compared with a human work?
  • How long do you think your conversational bot is mature?
  • What is the plan to mitigation if a conversational bot does not understand?
  • How do you mitigate high customer churn rate after deploying conversational bot?
  • Does your conversational bot have human-in-the-loop technique?
  • How do you transfrom from traditional conversation to an intelligent conversation by bot?

Read this article and download ROI Calculator for conversational bot: The ROI of (Conversational) AI

Business users do not care what is built behind the scene, today they care about how fast you respond to them, and how accurate your answer give them to solve their problems on services you are providing. Communication is key success to business, especially to customer experience management. If you can’t control the way you communicate with your customer, how much of confidence do you believe in your conversational bot?