Facebook’s Messenger updates chatbots with built-in language parsing, bot-to-human handoffs
Messenger adds natural language processing, opens up automated handoffs, streamlines native payments and extends customization features to desktop.
Facebook’s Messenger is looking to make chatbots smarter and is better equipping them to handle situations when they’re not smart enough.
On Thursday Messenger updated its toolset for bot makers to include: built-in language parsing, automated handoffs from bots to humans, a one-step payment process and better desktop support. It also rolled out new buttons for Facebook Pages to get people to chat with them on Messenger.
Built-in natural language processing
Bots are far from conversational savants. They may know what “hello” means but have a harder time understanding “whaddup.” They may also struggle to recognize when someone is telling them a date or location. To address this, Messenger is going to start automatically parsing some things for bots, instead of each bot maker having to think through all the ways a person may say something. Initially Messenger will use natural language processing to automatically detect and process nine types of things people may say — hello, bye, date and time, location, amount of money, phone number, email address and a URL address — and pass along that information as structured data to the bot.
Bot-to-human handoffs
In April Messenger introduced a way for developers to automatically have their chatbots handoff a conversation to a human. Now the messaging service is opening up that option globally and making it available through third-party firms like bot management company Assi.st and marketing management company Sprinklr.
One-step native payments
After adding a way for people to pay for products and services through their Messenger accounts in September 2016, Messenger is now streamlining its native payment system when making purchases on a brand’s site within Messenger, though keeping native payments U.S.-only during a continued testing phase. Through a new software development kit, native payments will become a one-step process for purchases made through the webview version of a business’s e-commerce site, as shown below.
Messenger is making native payments a one-step process for customers.
Improved desktop support
Messenger is aligning the desktop version of chatbots to be more on par with their mobile models. Specifically, by adding desktop support for Messenger Extensions — which are what bots use to do things like fetch someone’s user ID to tailor a conversation, and to customize the messages people can share from a bot’s in-app webview.
New “send to Messenger” buttons on Facebook Pages
Brands now have more options for the buttons featured on their Facebook Pages that people can click to start a conversation in Messenger. The five new call-to-action messages are: shop now, get support, get updates, play now, get started.
Matching phone numbers to Messenger accounts
For $99 bot makers can apply to send messages to people through Messenger based on those people’s phone numbers. This is basically a way for brands to extend their text-messaging sign-ups into Messenger. Even though there are 1.2 billion people using Messenger every month, not everyone in a brand’s phone book may be among them. Now developers will be able to find out if a phone number matches a Messenger account before sending a message, and they will also be able to access that person’s name through Messenger.
Chat Extensions for global Facebook Pages
Chat Extensions are Messenger’s way of letting bots into people’s conversations with their friends. However, since Chat Extensions rolled out in April, people conversing with friends in different regions weren’t able to use Chat Extensions for Messenger bots tied to global Facebook Pages. That’s fixed now.
Notifications for blocks, policy issues
Messenger will start notifying developers when their bots are blocked or run into an issue with one of Messenger’s policies, so that the developer can reconcile the problem.
Opinions expressed in this article are those of the guest author and not necessarily MarTech. Staff authors are listed here.
About The Author
