We bring good news from the frontier of human-computer interactions: the days of tedious hold music and incomprehensible telephones menus will soon be behind us.
The future of customer service and interaction will almost certainly be through chatbots on platforms like WhatsApp, Facebook and Telegram, using AI and automated responses to directing people to the information they need much faster than the annoyingly slow recording on the other end of the phone.
That’s the message coming from the first BotCon Africa conference down at 44 Stanley in Joburg, anyway, where developers in the field are gathered to learn and share best practices for building “conversational interfaces”, or bots to you and me.
Facebook’s Emeka Afigbo opened proceedings with an overview of how bots work on the Facebook Messenger platform. Love it or loathe it, most people who are coming online for the first time now experience the internet not as the web, but through the lens of Facebook or WhatsApp – and multilingual bots will be a big part of their future interactions.
So when Afigbo shares the learnings from the 33 000 or so bots currently active on Facebook right now, it’s worth listening.
During his presentation he singled out several bots for praise, including the flight information bot from KLM airlines and the IHG hotel group’s Messenger bot that will assist you to book rooms and order room service. ABSA got a shout out for its local banking bot, and in the promo space Call of Duty and Channel 4 (a TV channel in the UK) show Humans were also highlighted as examples of good practice.
For future bot designers, Afigbo says there are two main areas of best practice that separates good bots from bad.
The first is to manage the expectations of your users. Make it clear in the welcome page and first page what the bot can and can’t do. Make sure there’s a persistent menu to take them to information quickly if they know what they’re looking for.
The second, which follows on from that, is to limit “freefrom input”. The promise of bots is that someone can type any request and the bot will know what they’re talking about. Think more like a 1980s text based adventure designer than a Siri programmer, and your app will be more successful, however.
“If you give the impression that the user can type anything,” says Afigbo, “When they do that and the response is not what they expect, they feel the bot is not up to scratch.”
To help direct people to the right reply, Afigbo suggests using multiple choice “quick reply” responses rather than a text box (Is this the flight you want, Yes/No etc).
So there you have it. Go build bots.[Image – CC 2.0 by Simon Liu]