Chat bots: How talking to your apps became the next big thing

Apps that can mimic human conversations are one of the hottest technologies around right now. Here's why.
Written by Steve Ranger, Global News Director

Talk to the robot.

Getty Images/iStockphoto

Bots are applications that are designed to respond to conversational language. The aim is to create services -- whether that's the ability to order a pizza or to enter a meeting in a calendar -- where the dialogue with the app is as natural and apparently unscripted as an interaction you might have with a human.

Chat bots are like narrow versions of digital assistants like Apple's Siri, Amazon's Alexa or Google Assistant, designed to perform specific tasks. Interest in bots has rocketed recently and developers are racing to incorporate them into services built on popular messaging apps and websites to create a form of virtual customer services.

This has also sparked a mini technology arms race, as the big tech vendors try to persuade developers to use their bot platform. Google recently acquired API.ai, which has around 60,000 developers using its bot-building technology; Microsoft claims around 45,000 developers for its Bot Framework; and Facebook has around 34,000 working on its platform.

"So pretty much everyone today who's building applications, whether they be mobile apps or desktop apps or websites, will build bots as the new interface where you have a human dialogue interface versus menus of the past," predicted Microsoft CEO Satya Nadella earlier this year.

Some of the technology behind bots has been around for a long time, so why are bots so hot right now? That's largely because advances in AI technologies such as Natural Language Processing (NLP) and voice recognition are coming together to create a need breed of smarter bots that use AI to respond to questions in a much more natural way than was previously possible: for example, Amazon has launched a competition to build a bot that can chat "coherently and engagingly" with humans for 20 minutes about sports or celebrity gossip.

Add this to the rise of messenger applications like WhatsApp, and we could be seeing the emergence of a new type of user interface -- one controlled by conversation. And while this is going to be most visible in the consumer world at first, bots will also extend into the enterprise as mobile messaging is fast becoming a standard productivity tool alongside email and calendar.

For some companies, bots may become a way of supplementing (or replacing) human customer services staff; for others bots may create new ways of doing business.

A new bandwagon?

In many ways the bots craze mirrors what happened with mobile apps a few years ago, with lots of hype and over-enthusiastic developers jumping on the bandwagon.

"The excitement, however, is well-founded because chat bots do have enormous potential," said Raul Castanon-Martinez, senior analyst at 451 Research.

Despite the huge interest, many of the bots currently in use are relatively basic and more like an upgrade to the menu-driven (and often infuriating) interactive voice response services that have been around for years.

"We are nowhere near the expectation that chat bots have raised for AI and human-like conversations. There are important advances, but we're not there yet," said Castanon-Martinez.

Chat bots can fill gaps in mobile applications by making it easier to get work done (or just get that pizza ordered) in a way that's more natural on a smaller device: "Troops, for example, allows users to get work done by interacting with Salesforce over a very clean and simple UI -- Slack," said Castanon-Martinez .

"Bots are still in early days. We still have to see AI, machine learning and NLP advance to a degree where they enable natural conversations. The typical use case right now will be relatively simplistic yet powerful -- just not to the extent of a completely automated chat bot that emulates what a human can do. I expect we will get closer to this scenario in the next couple of years," he added.

Bank bots

Royal Bank of Scotland (RBS) is one company that is experimenting with chat bots for customer service using IBM Watson Conversation. The bank will begin using the chat bot with around 10 percent of RBS customers in Scotland to answer queries such as "How do I authorise my card to be used overseas?" or "How do I update my home address with the bank?"

This customer pilot follows a two-month trial of the technology earlier this year among 1,200 RBS and NatWest staff, mainly handling queries from small businesses customers with problems such as lost corporate cards or forgotten pins.

Chris Popple, managing director of digital for NatWest and RBS, said the aim was to make help and support easier for customers.

The chat bot is designed to respond to a customer query in a conversational way so that it can gather the information it needs to either answer the question or direct the customer to staff who can respond, which is better than simply giving customers a web page to read.

The bot will be able to respond to queries like "I don't recognize this transaction" or "I don't understand my balance" or "I've lost my card", which make up the majority of the questions asked, but as the pilot progresses the bank wants to the bot to handle harder queries.

"As we progress through the pilot and next year we want to get into more complex intents to start to test whether or not it can understand the underlying needs. So, 'I've run out of money' -- that could be a number of things, but first we need to nail the easier intents and as we go along we'll get into the more complex stuff," said Popple. If the pilot is a success the bot will also be rolled out to NatWest customers.

Building an effective bot requires a different set of skills to a standard IT project, including programmers, subject matter experts and dialogue writers. For the latter, the bank has turned to people who have written screenplays.

"The team is not very large -- under ten people -- but it's a mixture of those skillsets that come together," said Popple.

The aim is for the bot to have a personality, but not for it to pretend to be human.

"We're not in the business of making the technology pretend it's a real person, but we know now from testing it internally that people respond very well when it seems natural," said Popple.

In the future, the chat bot could also use IBM's Watson AlchemyLanguage capability to understand how a customer is feeling and then change its tone and actions accordingly.

"We want to sensitize it so that we can maybe pass you to a member of staff because it's an incredibly urgent thing, or change it to a slightly more formal tone if it's a more serious question. We don't think it will be very complicated, but we do want to have it react slightly differently depending on the tone and the urgency", Popple said.

More on chat bots

Editorial standards