X

The Mobile & Digital Assistant Blog covers the latest in mobile and conversational AI development and engagement

  • September 23, 2017

4 Concepts to Consider Before Building a Chatbot

Dan Brooks
Product Marketing Manager, AI, Bots & Mobile

 

What use is a chatbot without data, or the systems that house that data?  Not much, when you think about it. Sure, you could build a chatbot and hard code canned answers to users' questions into its programming, but if you want to build an AI-powered, intelligent bot that responds in a contextual and personalized way, you not only need natural language processing and machine learning tools, you also need data that the bot can access to provide its users with the most relevant answers.

In a new Forbes article, Alexa Morales speaks with Oracle Vice President of Product Management - Mobile, Bots & AI, Suhas Uliyar, to learn more about the larger questions that enterprises need to ask when developing chatbots.  In this context, "how do I build my first chatbot?" is too narrow in scope.  To survive in this hyper-competitive, on-demand world, large businesses must think in broader terms, across departments, technologies, channels, and even time if they are to avoid disruption from smaller more nimble competitors.  A more appropriate question that an enterprise might ask could be "how do I build an intelligent, AI-powered user experience that can connect with all of my back-end systems and adapt to different communication channels over time?" Answering this question still could lead a company to focus on chatbot development, but in the context of a larger, enterprise-wide, AI push that will be needed as user demands, and the personalized, automated responses created to engage with users, increase exponentially.

When considering these questions as you begin chatbot development, Uliyar advocates paying attention to four key concepts:

1.  Understand how complex each messaging channel actually is:

  • Every channel has different standards. Not only are different channels popular in different parts of the world, but each messaging platform has a different look and feel on the front-end, and on the back-end, operates differently.  To create a specific chatbot for each different channel sucks up valuable time and resources. A platform that lets a developer write code once and deploy it across multiple channels, however, offers tremendous business value.

2.  Don't trivialize Machine Learning and Natural Language Processing:

  • While you certainly can use an open source ML or NLP library for your chatbot, these tools have become "table stakes" for many of today's applications; everyone is using them. If you want your chatbot to stand out, you should consider allowing your technology platform to use its own algorithms to model both "supervised" and "unsupervised learning" so that your chatbot can make more accurate predictions than its peers.

3.  You need a good dialog engine to create a good chatbot:

  • If your bot can't understand that "What's my balance?" and "How much money do I have?" are two variations of the same question, you're going to end up with a lot of frustrated users.

4.  A chatbot without enterprise integration isn't tapping into its full potential:

  • As stated above, what's the use of creating a chatbot when it can't connect to your back-end systems to access all of that data, in a personalized way, for your customers?

Fortunately, Uliyar reveals that Oracle Mobile Cloud Enterprise can address all of these concepts.  To read the full article on Forbes, click here, and to learn more about Oracle Intelligent Bots, visit oracle.com/bots.

 

Be the first to comment

Comments ( 0 )
Please enter your name.Please provide a valid email address.Please enter a comment.CAPTCHA challenge response provided was incorrect. Please try again.