Chatbots are everywhere.
If you’ve ever talked to the robot voice in your phone or received a text message from a restaurant, you’ve used a chatbot. They’re quickly becoming the new hot trend in tech, and for some very good reasons. Over the last 2 years, messaging apps have far surpassed social networks in overall monthly active users. What this means for brands is that the landscape of customer interaction online is changing. However, conversational interaction with systems has been around for a long time. This is not to say that they aren’t improving with modern AI – they are, but the general concept has existed since the dawn of modern computing. ELIZA was one of the first chatbots, developed in 1966 by Joseph Weizenbaum at MIT. Throughout the early years of the internet, chatbots enabled you to play games in AOL chatrooms and download files on your local BBS. Today they do a myriad of things, from finding reviews for good Chinese food nearby to fielding customer service requests on a website.
Chatbots in the wild
There are a lot of chatbots out in the world -- after all, there’s a good chance that you carry one around in your pocket. Amazon is even running a contest to spur developers, but here are a few that we found interesting:
- Taco Bell has a chatbot that allows Slack users to order Tacos directly from chat.
- Nordstrom’s has a bot that helps you find the right gift for that special someone via text messages.
- Uber has a Facebook Messenger bot that allows you to order a ride and invite your friends.
How Do They Work?
There are few components to the modern chatbot. These consist of a data source, an interaction layer, a logic layer, and optionally AI. This sounds fancy, but the secret sauce behind a good chatbot is content and conversation design, not the underlying engine. The data layer is typically a website database or a collection of APIs. The interaction layer is where the bot interacts with your users – a text message, a chat with Siri, or Facebook Messenger for example. The logic layer is a simple machine, with the base engine searching for keywords and phrases to inform the software of the user’s intent. Artificial Intelligence can take these systems to the next level, however. When you give a chatbot a prompt such as “I’d like to order a pizza”, you might expect a response like “What type of pizza would you like to order?”. Through utilizing an Artificial Intelligence layer and a good data set, you might make your system respond with personalized information instead: “Hey Jeff, how about your regular 16” Supreme?” AI also opens the door for better recommendations, voice input, and sentence recognition. With AI the potential exists to remove the multi-step questioning process (what product, where is it going, how to pay for it) and make compound requests to internal systems much easier.
How Do I Build One?
Currently, there is a great chatbot Drupal 8 Module released by the White House. (Yes, that White House.) It’s been forked and improved from their original spec, but this is a good start for people wishing to do their own development and hosting for a chatbot. For those wishing to get up and running with less technical rigmarole, Amazon’s LEX service is a great option to jump in and start working on conversational logic right away. The largest messaging platforms all have their own APIs to integrate with, as do services like Siri on iOS. Amazon, IBM, and Microsoft all have public SaaS services that enable affordable AI for your applications. There are also services like Twilio that will enable you to text message users if you like. With a little ingenuity and some elbow grease, making a chatbot is easier than ever before.
Technology aside, chatbots have great benefits for businesses. If built properly, they can both greatly increase customer satisfaction and customer support availability while simultaneously decreasing the costs associated. While chatbots are not yet sophisticated enough to fully replace a true customer support team, it is only a matter of time before most customers interact exclusively with a bot that represents a company. You can already see this change starting to happen with self-checkouts at your local supermarket.
A New Dawn
With the advent of chatbots being the face of companies in a new era of business, there are several concerns:
Could chatbots be a security risk?
If your bot can complete too many tasks without verifying information, or if your conversations are setup in a way that can be exploited, chatbots could pose a potential security risk. Without voice recognition or some other form of verification, conversation designers need to be aware of the implications of exposing sensitive data.
Too Big Brother-y
With too much information and related metadata exposed to the engine, users may feel wary of using your software. The whole idea behind a chatbot is to make interaction better for the user – creeping them out may not be the best approach to client retention.
They Aren’t Humans
Artificial Intelligence is helping to bridge this gap, but not every sentence structure or phrase will work with a modern chatbot. They also get hung up on slang terms or phrases, and have difficulty interpreting emoji. Chatbots aren’t new, but they are seeing a rise in popularity as the interface between businesses and customers changes online. With the long list of the benefits they can bring to your brand, you should start thinking about how you can leverage chatbots to enable success.