Kik launched their bot store and opened their API to the public. In order for your bot to be published in the store, it needs to be reviewed manually, same as in Slack. From the developer perspective: very neat, they took the simplicity of Telegram's BotFather and then simplified it even further. Offtopic note: I love Kik codes and how fast they are scanned. Yeah, I know it’s uncool to even say the word “QR”, but I have hope in these new QRs.
And then, F8 is coming out this Tuesday. The Message section of a Facebook business page is going to become relevant again. They are going to push for this a lot. Ah! Messenger codes were released, too.
What’s keeping me busy, other than writing this newsletter, is building a chat bot platform for businesses to create interactive conversations visually. We’re staying under the radar with a small number of partners that are creating working prototypes in a few hours and deploying in production in a few weeks. If you want to move fast and launch across Messenger, Kik, SMS and Telegram, ping me and let’s talk!
Even if it didn’t happen yet, it kind of feels like old news already ;). There are other news: a new section in Messenger that suggests businesses to chat with, short URLs like m.me/my_username to chat to businesses, and scannable Messenger codes. It is very clear that developers, businesses and publishers are excited. Now it’s time to get the ball rolling and see how consumers actually interact.
Insightful article about chat bots from the perspective of brands. It ranges from "bots will be the new websites" by Kik's founder to a "We use real people to engage with our customers" from Tesco operations manager. It makes sense, we never said chat bots would apply in every single scenario.
LINE, the leading messaging app in Japan with 200 MAUs, is putting a lot of effort into business to consumer interactions. On top of the chat SDK, they plan to offer something similar to wit.ai and LINE Beacons. Also, they’re working on deep web integration, to be able to login and pay in a website from LINE.
Some notes about how to align the expectations of the user with what your bot can do. Who initiates the conversation is key. In short: if it’s the human, then you have a hard problem to solve. Also, you want to narrow down the domain of your bot as much as possible, and communicate that clearly to your user.
First post in a series about deep learning, that lays out the difference between retrieval-based models (pre-defined set of responses) and generative models (much, much harder), and goes into the challenges of natural conversations, including links to very recent research papers. The following posts in this series will get very technical, so be sure to check back if you are serious about deep learning for chat bots.