The Beginning of the Bot: Separating the 'Bot' from the 'Messenger'

Like so many of my peers in the AI tech space, I'm sure, I was eagerly watching the stream of Facebook's F8 conference yesterday and particularly the presentations by Frerk-Malte Feller and Jessica Lee of Facebook Messenger.

If you've missed the buzz in both tech and mainstream press, heralding the arrival of the Bot on Facebook Messenger, here is the crux:

• Messenger is now accessible to developers via APIs, meaning third party content can now be delivered into Messenger in the form of 'dialogue'
• A chat bot platform has been included in the release, enabling developers to automate conversations using templates and other content structures

During the presentation we were introduced to some of the pioneering bots on the platform, like those from online retailer Spring (impressive) and CNN.

While it's early days for these automated experiences,  what struck me about the demonstration of these pioneering bots was the limited use of natural language recognition and the transactional nature of the bots, and this brings me to the subject of this post.

Last night I got a message from a friend (not on Messenger) asking me whether this development kills my business, Twyla - our product is an AI customer support chatbot platform that learns from human conversations and can thus deliver human-like support via Messenger, amongst other chat channels.  On the contrary, I told him, the Messenger development is great news for Twyla because it means more opportunity for us to expand and push our vision of chatbots as the medium of the future for customer support.

His misunderstanding stemmed from having a superficial understanding of what a bot really is.  He thought that Messenger was somehow providing the AI firepower that any bot needs in order to work.  It's not, not yet anyway.  Facebook has acquired Wit.ai, a kind of DIY bot creation platform which it may integrate more closely with the Messenger product itself in time.  And while Facebook is working on it's own AI-powered assistant, M, that is not what the Messenger update is all about.  It's about opening up the interface of Messenger to receive messages from sources outside of its own ecosystem.  It brings the audience to the bots, basically, not the other way around.

So the bottom line here is: Bots on Messenger are not Messenger bots.  They are simply bots that are using Messenger as an interface layer to reach an audience.  The underlying technology required for powering those bots has to sit in another system, which repurposes data into the format that Messenger allows.

Take for example KLM's impressive bot.  It can display your boarding card right there in the conversation.  The card content itself is not living in Facebook any more than the passenger's booking data is.  There is, instead, an intermediary technology required to deliver that data into Messenger, in a format that Messenger likes.  And to deliver that data in a conversation, artificial intelligence is required from that intermediary technology too.

So, with that in mind, when building a bot, you're not limited to Messenger as an interface for that experience.  Because the AI technology - the bot itself - lives in an independent system, you can also facilitate the same conversations with customers using other interfaces, like Telegram, WeChat, Slack, web chat or - in the near future, we hope - WhatsApp.

These independent systems are going to be at the centre of a gold rush now, which Twyla and its contemporaries are a part of.  The bigger challenges of natural language and how to deliver deep, contextually specific content - like customer support - loom large.  But this announcement by Facebook, and the accompanying hoopla - however confusing -  are welcome steps towards that brave new world ;)

https://twylahelps.com