“Chatbots” are very popular. What is it ? It’s a mix between “robot” and “chat” : chatbots are conversational robots that interact with users through an interface that often look likes instant messaging.
Chatbots are the future of customer relationship and employee relationship.
Today the most popular example, the one all businesses are looking at is Facebook Messenger’s chatbots that allow brands to interact with customers and offer them different kind of services, in the same wat KLM allows passengers to do with messengers everything they used to do on their site or on the mobile app to manage bookings.
In China, Wechat has become more than a messaging tool : conversational interactions with brands for any kind of service is the new normal there and there is no reason why the same won’t happen here.
Some brands or pure players also propose chatbots services via SMSn as I already explained for travel personal assistants.
The interaction can also happen through a search engine, like what The Northface did with IBM Watson, delivering one of the first examples of cognitive commerce.
But the future of chatbots is not only on the customer side. In a future post I’ll elaborate on the idea of HRbots (HR Services for employees) and robots supporting collaboration. No matter the field, bots will be a part of the future of the digital work environment and of employee experience.
A more or less artificial intelligence
How does a chatbot work ? To oversimplify : you need a user interface (sms, messenger or other), that will capture the message and display the answer, it’s the front end, andÂ backing consisting of a machine that will process the request. Both are Independent, you can use Facebook Messenger as a front office and IBM Watson as a back office.
How to build a minimalist chatbot, an example with IBM Watson.
That’s the point where things get more complex and where it’s important to know one’s purpose and kind of engine the chatbot needs.
On the user side there’s a dialogue with a machine so most users think it’s an artificial intelligence engine. In fact, sometimes, the intelligence is so artificial that it’s not existing.
There are two ways of powering a chatbot.
1Â°) Artificial intelligence
There’s a true artificial intelligence behind the chatbot, like an IBM Watson. No matter what you say it will try to undersand and give an more or less relevant answer. Over time the machine will learn and improve (cognitive computing). The main constraint when you launch such a projet : you need lots of data and time to teach the robot.
2Â°) Rules Engine
In this case there is no intelligence. A rules engined will be programmed to give a given answer to a given question. The problem is that the user asks something that as not been anticipated the bot does not answer and the user is frustrated. Contrary to a cognitive approach, the robot does not learn. If you change your products you have to add a rule the engine will learn by itself. It may become very complicated if you bot has a wide range of interactions and matters to deal with but it can be started fast and is less costly.
An hybrid approach may make sense : a rules engines to start while the AI is learning.
So, the same name covers two really distinct approaches. They don’t cover the same needs, ambitions and each one works very well for a given need. But before you start such a project you must be aware of the differences and don’t buy a nice and cheap solution expecting to get the Rolls of AI.
Photo credit : Fotolia