Once upon a time, on a blue planet, marketers were given a precious gift by the Goddess of artificial intelligence;
chatbot. These nice computer programs were able to automate tasks, by
chatting with Muggles through a conversational interface. Thanks to AI,
they can answer to complex requests, create personalised answers, and
more importantly, they can learn, and the interactions can improve over
time.
Do you need to order online a new magic wound? Easy! No need anymore
to scroll through lots of webpages on a website. Just use the
conversational interface and let the nice little avatar helping you. It
will ask you questions, and your answers will help the program to
satisfy your needs.
When
looking at the picture below, it is no surprise to see how popular
chatbots are becoming. For marketers, it’s a magical tool to engage users and grow their business. But what is the share of interactions handled by AI? In 2018, it was only 5%. But, by 2022, this figure is expected to increase fourfold to 20%.
 |
| Retrieved from SocialMediaToday |
Is it just a fad?
Definitely no!
In addition to the convenience it brings to both companies and Muggles,
sorry, customers, as mentioned above, chatbots bring tones of
opportunities. No need to download additional apps, chatbots can be
integrated into websites but also in messaging apps, the place where
customers already spend their time. Still not convinced? Read that: in 2016, for the first time in history, people were using messaging apps more than social networks, they had over 5 billion monthly active users!
It is also important to note that chatbots are part of a larger shift we are seeing in consumer behavior.
When they research something, users want the information they need as
quickly as possible. To do so, they are turning more and more to voice
search and other technological advancements. They want to see value in
the places they already are. Moreover, some people even tend to prefer
to talk to robots about personal matters instead of humans.
Have a look
at the video below that talks about Replika, a popular app (7 million
users!) that allows people to share their feelings with an AI companion.
With magical powers came great responsibilities
A chatbot is a magical tool. However, it should be used cautiously. When studying the performance of chatbot, some researchers argued that trust is the focal point of successful interactions.
One of the challenges is the double-edged sword of messaging.
People are accustomed to having complex conversations with other humans
on messaging apps. It can be really frustrating to chat with a machine
that is completely off the mark. Have a look below at my personal
example (zoom in for more clarity). One day I got some issues when paying for a module to my British university through my Revolut
account. I tried the chat at 7 am, and the chatbot did not understand
my request at all. Hopefully, when I asked to talk to “someone”, I was
positively surprised about the promptness of the banking app. After just
a few minutes, my problem was solved by a nice agent. Even though the machine couldn’t help me, the system quickly recognized its limits. In my opinion, Revolut’s
strategy is right. The machine is here to filter the requests, but if
it’s too complex, an agent takes over. And according to Susannah
Richardson, in her article for IndustryWeek, “the
real hallmark of success for an automated self-service system is its
ability to understand cases where an automated response is not
appropriate”. The main challenge lies here. The system needs to recognise when a human intervention is needed. My case was pretty simple, as I asked for it.
 |
| Personal conversation on the app Revolut |
My
situation shows how challenging chatbot can be. But there are also
other risks. With e-mails, if marketers spam someone, technology can
help by filtering spams. But with bots that are functioning inside
messaging apps,
spammers can pollute personal space. And this is not very pleasant. I
am not really looking forward to having my WhatsApp looking like my junk
mail folder.
Another
challenge is that a lot of background data is required for an AI
chatbot to be able to answer to requests. Twenty customers may ask for
the same thing but in twenty completely different ways. For example, a
credit card company analysed how customers asked for their current account balance. For such a simple request, they did ask it in 2100 distinct ways.
