The Techology

​

What are Chatbots? In essence, they are systems that allow individuals to converse with them using natural language. On websites in text form, website ‘chats’, and in voice form typically via automated calls.

 

The technology is quite multifaceted but, there are basically two types:

 

Rule-Based – These bots are programmed with a predetermined set of rules. Think of it as a conditional if:then process or root and branch structure. As you can imagine it works well for simple clear tasks where input questions and output answers are clearly defined. The ‘bot’ itself is stupid and only as good as the training provided to it.

 

Despite this simplicity, there are still challenges with different languages and question styles. This can keep developers busy as they attempt to program in a multitude of different permutations. For example, a user asking for a bank statement could say ‘Can I have my statement?’ or ‘what is my latest statement?’

 

They also need to be audited to ensure they are processing requests correctly after all it is a computer program, and the rule scripts need to be kept updated to reflect current processes and business rules.

 

AI Bot: Using complex algorithms and training, typically using recordings of real-life client calls to mimic actual conversations, this application is significantly more complex than the rule-based approach but, when done properly, can produce bot’s whose interactions can be much closer to that of a real person. They are often referred to as smart bot’s as they are self-learning versus the earlier rule-based model. The new Emotibot’s push the boundaries even further as they develop an understanding and can respond effectively to human emotions.

​

These AI Bot’s introduce a whole new set of issues for organisations ranging from - the validity of the training dataset - the ethics of the trained AI's - to the quality of decision making, requiring continual auditing to ensure the AI learning algorithms continue to work effectively.