Conversational Interfaces Need a Different Content Management System

It’s been more than a year since I began working on conversational interfaces (chatbots, to be more specific). My team and I have built bots for Slack, Twitter, and of course, Facebook Messenger.

A common thread in all of these bots is content. Content is the key to a successful bot. It plays a role at many levels — ranging from the bot-dialogue (how the bot speaks to you) to the actual content delivered from within the bot that the user consumes. At the same time, there is a unique set of restrictions / needs which bot content demands.

Today, in the version 1.0 of the bot industry, bot creators are using legacy tools and repurposing or retrofitting legacy content into bots. This is a key reason why bots are struggling or making users feel disappointed.

Well designed bots are, in reality well designed conversations.

Writing Native Content For Bots

Content that’s designed for bots must have:

• Short and precise content bites that are quick to consume
• Emojis for meaningful, visual attention
• A high balance of images and gifs throughout

These are unique requirements for content creators at many levels. For some time to come, while AI is yet to eat up all the editorial and content writing roles, most of this content will be created by humans.

The challenge is that today’s writers are using yesterday’s tools for creating content for tommorows interfaces.

A Bot-Aware CMS

So what would be an ideal way to create bot content? Let’s take a look.

1. A wsyiwig editor designed for writing short pieces of content, in built emoji suggestion, and content driven image or gif suggestions.
2. A way to visualise the content as it is created in various bot platform interfaces and a smart way to adapt it to different needs — a “write once, read anywhere” format.
3. Text analysis at runtime which helps writers ensure their content style is conversational.
4. A speech to text engine in built for the ability to account for voice bots.

While it might seem simple to hook up an interface or a wsyiwig editor with all these pieces, the real mettle of a true Bot CMS would be the magic sauce of NLP.

Train the Brain

A bot-aware CMS should be capable of training the intent engine of the bot in real time, as writers are creating content. It’s common knowledge that the capability of a bot to understand human conversation on a topic comes from intent-based training. There is a whole array of intent engines out there — Wit, Recast, Rasa — but the real magic of intent engines is in the training models and the real value of models only comes from the training data set. A CMS that is truly future proof for the conversational world we are heading towards would have this capability built into itself.

I am actively researching this area and figuring out what else a CMS of the futuee would need. I would love to hear ideas, thoughts and counter thoughts to this premise. And if you want to collaborate or get more details on my work DM me on twitter or Facebook.