Tinka as Facebook Chatbot
Why the charming digital assistant can now also be found in Facebook Messenger
For two years, T-Mobile.at has been offering the digital assistant Tinka on its website. She helps customers with questions about products, prices, and contracts, solving many issues directly in a chat-like dialogue.
When Facebook opened its Messenger platform to developers in April 2016, among the things to emerge was a Facebook version of Tinka. She is integrated within the new Facebook Messenger chatbot interface. Customers can now contact customer support directly, using the Messenger app that they already use on a daily basis. However, even if a problem cannot (yet) be solved fully automatically, Tinka will not leave them in the lurch. The chat session is seamlessly transferred to a customer service employee; in many cases, the customer does not even need to leave the familiar interface. This customer-friendly solution not only makes T-Mobile Austria more easily reachable, it also ensures that service is faster and better.
Why did T-Mobile Austria choose in favour of this particular social media platform? Mostly because it offers huge potential for customer interaction using their preferred medium and it provides them with service where they most want to receive it.
The numbers speak for themselves:
- Facebook Messenger has more than 900 million active monthly users.
- Every month, more than one billion messages are sent between users and business pages via the Messenger app.
- More than 2.5 billion users have at least one Messenger app installed.
- It is predicted that in the coming years, around 3.6 billion users will have installed a Messenger app.
- 49 per cent of smartphone users between 18 and 29 use messenger apps.
In business terms, these numbers equate to a tremendous reach. Tinka, as the T-Mobile Austria Facebook bot, makes conversational customer care automation possible: in other words, she provides automated customer care in dialogue form.
Unlike most static Facebook bots, Tinka uses keyword-based application cases and processes that have been defined within the Deutsche Telekom international AI project.
Most of the other FB bots to date generally have been developed using tools, such as WIT.AI. These define a static, tree-based information flow, supplied using defined Q&A texts that the customer needs to consider to (hopefully) receive a successful response to his or her question. The maintenance of this application method tends to be very static and labour-intensive.
In contrast, the design of Tinka achieves the following:
- Ensuring consistency and accuracy of available content for both the web-based and Messenger-based Tinka – using a single backend
- Dynamic content updates for Tinka: If Tinka’s knowledge database is updated, this information becomes available to Tinka directly in Messenger through APIs
- Consistent user experience: The virtual assistant on the T-Mobile website allows for the limitations of the Facebook messaging platform when preparing content for the Messenger solution
- LiveAgent handover: This enables the Messenger version of Tinka to transfer each customer interaction to a human customer service agent, directly through the platform. This can be triggered via certain pre-configured actions, such as Tinka not being able to satisfactorily answer a question, through a keyword or by using a static menu option. When Tinka transfers a customer directly to an Austrian customer service agent, the agent can instantly view the chat history and help the customer without having to request information again.
Tinka is one element of the much larger eLIZA project.
Further key features of Tinka in Messenger include:
- Users can switch between smartphone, desktop browser, or native app while chatting with Tinka, without losing the dialogue history. In other words, all user information is preserved.
- A secure browser session is seamlessly launched (the customer needs to log in to My T-Mobile) if private information has to be transmitted outside Facebook Messenger. This ensures the necessary security is maintained.
- Support of all four main types of interaction – pure text, URL buttons, postback buttons and static menu
- Support for the Messenger app toolbar at the bottom of the screen. This allows users and Tinka to use various functions such as, “like”, “send image”, “location”, and so on. Here, the user can add emoticons to express feelings – all using a format that contributes to the machine learning of the AI. Currently, this is being developed as part of the larger NG eCare project.
- Integration within the social media platform for T-Mobile Austria service employees. This allows tickets to be created, which are necessary to forward requests – originating from Tinka interaction – to customer service. This happens seamlessly and includes the forwarding of chat history to the customer service agent, who can then carry on with the dialogue.
Such an approach makes T-Mobile Austria’s customer service distinctly more flexible and offers the following advantages:
- As automated customer care on Facebook, Tinka solves recurring requests autonomously. This allows the human colleagues in customer service to focus on more complex requests, and allows them more time for better-quality and more complex interactions with customers.
- The first point of contact, Tinka can also solve problems that occur outside normal opening hours. More demanding requests are forwarded to service employees using the social media platform, allowing them to be processed the next working day.