Facebook bots: How to reduce 40% of human service’s demand?



Facebook bots: How to reduce 40% of human service’s demand?

The creation of a bot doesn’t end when it’s launched. During the year of 2016, E.life was responsible for developing 40 chatbot prototypes for companies, but only after the first project went live we could really understand how the application would affect the Social CRM operation.

 “Bots are applications (robots) created to simulate human interactions with real people.”

The perfecting process of a bot also exists, just like in any other project. In our case, there was a significant increase of robot intelligence because of feedback (addition of new answers, terms, questions, etc), that are essential to guarantee service efficiency. But before we talk about our experience, it’s important that you know there is more than one type of bot in the market.

Types of bot

●      Service bot: quick access, practical and, specially, dynamic towards most questions and answers from the brand.

●      Concierge bot: service request, such as bank statement, account balance or even a taxi request or hotel reservation.

●      Shopping Bot: shopping directly through inbox.

With all this support, not even the “owl-client”, the one that relates to brands during night time, will be unanswered, even if it’s human team isn’t operating.

Below, we shared the main quantitative learnings we had with this service’s implementation.


In numbers

During the bot implementation week, we noticed a 50% growth on volume of mentions through inbox. 


Even though the volume of mentions has grown because of the start of the bot, the number of unique users seeking inbox services remained stable during that time, which means that installing the tool stimulated conversations and allowed the brand’s client to go through a universe of information about the services that, until then, he probably didn’t know. 



Until the moment of that analysis, the impression we had was that the bot “boosted” the page and gave his service team more work. However, we also identified a 40% reduction of natural demand that, usually, would be answered by the Social CRM team, usually concentrated on “Information” or “Requests”.


To reach this result of 40% reduction of human demand, we studied the real volume that was received by the page so we could understand the application’s potential of problem solving according to the messages’ nature. The process consists of:

1.     Mapping the volume of mentions received through inbox.

2.     Classifying the nature of the services (complaints, information, requests and compliments).

3.     Expanding the amount of subjects inside each nature.

4.     Defining the bot’s autonomy level to solve these subjects accordingly, with the possibilities of the application’s integration with other tools of the company.

5.     Clustering many question possibilities for the same answer.

6.     Producing creative content to guarantee the dynamics of the tool (gifs, images and infographics).


Brands that have already joined

In Brazil, some brands have already started using the new technology:

  • Vivo (Questions and answers): automatized interaction system through a menu in which the user navigates through the different options that lead to the website.
  • DOTZ (Questions and answers + artificial intelligence): interaction system that, besides allowing navigation through a menu, also comprehends random texts that the user types on inbox.
  • Banco Original (Artificial Intelligence + integration with data base): in this application, instead of directing the user to finish some transactions on the website, by integrating CRM with the bank it’s possible, for example, to request account balances, statements and bills through inbox.


How to measure and follow on real time

So you’re not in the dark, Buzzmonitor is ready to follow up and automatically classify every trade by consumers, so you can follow the most searched for subjects via dashboard.





Request a live Demo


leave a comment