Friendly bot answering your helpdesk questions? A hip shopping assistant who knows both your style and the latest trends? An assistant who knows which latte you prefer on your way to work? Look no further, the bots are here!
Why bots are appealing for business?
Companies want to be where the customers are. About a decade ago the place was Facebook. Today, the messaging platforms like Snapchat, Instragram, WhatsApp, and WeChat have already passed the traditional social media sites in the amount of time spent at the service (At least for the millennials. Grandparents still use Facebook). The only traditional service that has kept up with the youngsters seems to be YouTube.
People are extremely curious about artificial intelligence (AI). Personal digital assistants like Siri, Cortana and Google Assistant have introduced the concept of bots and conversational platform to the masses, and they set the bar height on what people expect (and hope) from the bots. Even if the bots are most of the time annoyingly stupid, they kind of provide a glimpse of the future.
I like Cortana reminding me to pick up my kids from their hobbies without me explicitly setting up calendar alarms for this. It’s just cool. As if the bot actually understood my needs.
Bots have one distinctive feature that no other platform has: You can build apps with an attitude. You can use this to give your brand a personality. Bots also don’t require installing your brand app. Your brand bot appears as a normal contact in your messaging app contact list – just like your real beloved (human) friends.
Seeing is understanding, so I’ve included a lots of video links to this blog article. The WeChat promotional video below is an excellent overview on why & how companies can start their journey on the new platform.
Video: WeChat for Business & Developers
Most of the existing bots tend to fall into a few main categories, which we explore next. Check out the videos behind the links!
H&M & Sephora Kik Bots are typical examples of the shopping bot category. The approach fits most shopping scenarios, which offer an alternative to (or compliment) existing web stores. These bots also introduce a key technique used on controlling the flow of conversation: The On-Bot dialogs. The dialogs are used to avoid the need for natural language understanding, which is still a challenge for computers. Dialogs also make clear what the bot expects from the user, what the bot can (and can’t) do, and guides the user to the scenarios bot knows how to handle.
Nike, McDonalds and Coca Cola traditionally invest heavily on branding. Brand Bots range from passive “get the latest news and offers” to very active campaign bots like in the Coca Cola and McDonalds examples. Competitions, events, fireworks, unicorns and glitter – share your golden moment with the brand.
Personal Digital Assistants
The Personal Digital Assistants (PDA) are the cutting edge in bot AI. They use voice and natural language for communicating, and sometimes even come with an actual home device like Amazon Echo. The most widely known PDAs are Apple’s Siri, Amazon Alexa, Microsoft Cortana and Google Assistant. Building this amount of wisdom in a bot usually requires more cash than an average company (or a small country) has in it’s disposal, so I recommend not trying to write your own generally intelligent assistant from scratch. Instead tap on the programming APIs these PDA’s use internally (e.g. Microsoft Cognitive Services), or aim to integrate your own intelligent agent as part of the parent PDA skill set.
One Trick Bots
One trick bot solves a very specific and simple problem like order a cab, reserve a train ticket, reorder your favourite cup of coffee or translate this text. One trick bots are usually quick to make, and may actually produce real value to it’s user in it’s own narrow field. Most bots start as such, and may later grow when more features are added.
The near future?
I would still call most of the bots research projects, as very few make significant commercial success. The biggest obsctacles are natural language processing (NLP), and the lack of training material needed to train even moderately intelligent AI’s…. The bots just don’t get you yet, but they are constantly improving. These prototypes are a necessary step to have real experience on human-machine conversations to improve the NLP AI and the skills bots have. Many companies like Microsoft has published conversational AI’s like Xiaoice to get this real world experience (=data) to train the AI’s with.
Should I build a bot?
Yes, but check that your expectations are realistic. You are not writing a bot that passes the Turing test. The examples in this article should give you a fair understanding about current technology capabilities. There are plenty of bot building frameworks to help you start building your own bots. The fun part is that the job is slightly different than you have used to in typical IT projects: You’ll likely spend most of your time working on texts; writing phrases that bot uses to reply to user (bot output and potential persona), and reading hundreds of different phrases the bot users have typed in to order coffee and translate them to intentions (bot input).
I’ll write an article on how to design a bot later on. Until then: Happy experimenting!