As Facebook presented extensive features of the Messenger bots in the F8 conference in April 2016, not everyone understood the impact that would result in communication between companies and their customers.
The days that followed saw us meet internally and discuss the associated potential and technical requirements. My colleague, Daniel Brolli, and I would like present an overview of the opportunities and conditions that are associated with Messenger bots.
Current statistics show that a large number of customer requests has moved from the traditional public pin boards to Facebook. Many customers are now using the private message function on Facebook pages to communicate with companies.
This means: All signs indicate that Facebook Messenger will soon be a central element of communication between companies and their customers. This is probably what Facebook also wants. And the Messenger bots will hereby be critical.
The fact is that similar bots, the so-called chat bots, have been around for some time and partially proven successful (e.g., Telegram for Messenger).
A bot (also called chat bot or Messenger bot) is a type of a digital assistant that responds to user requests based on predefined/standardized or individualized/personalized answers. The bot hereby relies on diverse data sources: The user’s profile data (e.g., current location, name, gender, etc.) and the customer database if linked with the profile or request (e.g., customer data or data from a specific order from an online shop).
The bots are meant to simplify and speed up the processing of customer requests or shopping and service processes. They should help get rid of emails, hotlines, or onsite chats with support employees or consultants. Here, bots need not necessarily replace all human workers. They are only meant to filter out or respond to simple requests quickly and effectively. They should replace emails and phone calls as the primary point of contact for customer requests and should only fall back to personnel if necessary.
This could have an impact on smartphone apps and web shops: If people become aware of the benefits of an alpha app/all-in-one system, the willingness to download an app from a specific shop operator or web shop could even go down further.
The advantage of Messenger bots lies in the fact that you have everything in a single app. Whereas people still use multiple platforms and communication tools to search for products (e.g., via Google), order from a specific online shop (e.g., Amazon), receive order confirmations and receipts via email, and then contact customer support through email or by phone, Facebook Messenger could soon enable you to do all this in a single app.
Linguistic peculiarities (dialects, spelling errors, etc.), misinterpretation of customer requests, wrong responses, or even failure of the bot to respond to customer requests might lead to unsatisfactory user experiences. Facebook certainly wants to avoid this, particularly through AI (artificial intelligence).
Facebook uses "Deep Text" in order to better understand the requests. This basically entails machine comprehension of text. The current technology is able to comprehend and interpret 1000 different texts in over 20 languages per second. It hereby tries to identify the intention behind the statements in the text.
Example: If you text "I just boarded a taxi", you are definitely not looking for a taxi, but someone who writes "In need of a taxi" certainly is.
Messenger bots can therefore be close-knit and use predefined question-answer combinations to generate an appropriate response, or try to use AI to understand the individual needs of the communication partner.
Since bots can also provide proactive information in form of a subscription (e.g., Poncho’s weather bot, promotional bots, etc.) such information should be as relevant and personalized as possible to make sure bot responses are not annoying or quickly considered spam. Here, it is not just Facebook who needs to optimize the bots. Developers and publishers also have a bit to do (also from the insights point of view).
Figure 1: Messenger showing information from an airline about the booked flight (Source: Facebook Developers)
Several companies are already working feverishly on the design and development of chat bots for diverse platforms, especially the Facebook Messenger. The Dutch airline, KLM, is known as a pioneer in the use of social media and is currently using Facebook Messenger not just for customer requests but also to communicate information and data about booked flights and current changes.
You can find a list of Facebook Messenger bots under this link.
Figure 2: Food2Go Messenger bot (Source: Facebook Deverlopers)
Figure 3: SmartWheather Messenger bot (Source: Facebook Deverlopers)
The functional principle of a bot is rather simple. Facebook has hereby done a lot to ensure that the development process is as easy and transparent as possible.
A bot requires:
Once you have met these conditions and have successfully linked your bot to a web server in the initial connection (see the Getting Started Guide from Facebook), you can go ahead and code the actual intelligence of your FB bot.
Here, there are no limits for developers. Both in terms of the programming language (Java, nodeJS, PHP, …) – since you can have everything on your own server – as well as the bot intelligence, which you are free to program as you wish. Basically, a chat process always proceeds in the same way.
The user must first contact the company (i.e. the bot) via Messenger and start the conversation. Facebook hereby offers different ways and means of doing this.
Similar to the Like buttons, you can integrate a link on your website that either leads users directly to the Messenger app or to the desktop version of Messenger. You can also allow your users to search for the desired bot in Facebook or Messenger and immediately send their request.
Here, the user sends a message to the linked Facebook page. The message is then forwarded to the registered web server and interpreted based on additional meta data e.g., User ID, message text, MIME type of the text, etc. The server then responds to the message depending on the specifications.
The bot either responds with a simple message, a web service activated, or the incoming text parsed and reviewed based on syntax and context enabling the bot reply with an appropriate message depending on the interpretations. With the latter, you can develop a "smart" bot since such bots understand what the user writes and are therefore able to provide a more appropriate response.
Just like others are doing, we too are working feverishly on initial implementations and applications of Messenger bots since we believe that they simplify communication between companies and their customers. We strongly believe that Messenger bots will soon establish themselves among customers and companies. As much as the advantages are clear, the quality standards must still remain in the background. And this is exactly where developers, companies, and Facebook are needed.
Published on 08/03/2016 by Karim-Patrick Bannour.
Karim-Patrick Bannour is founder and CEO of the social media agency viermalvier.at. Together with Anne Grabs, he wrote the social marketing best-seller “Follow me! – Erfolgreiches Social Media Marketing” (Galileo Computing). He also frequently speaks and gives workshops for organisations such as ÖHV or The Austrian Federal Economic Chamber. In 2016, he founded the Amazon agency MarktPlatz1 and together with his team supports providers and traders on Amazon.