How to build an Agent

jogoterbaire@

For those who want to try to build their own personal agent, let's try to briefly explain what the fundamental steps are:Needs analysis: in this phase all the possible requests that the agent must respond to are highlighted and the available resources are identified (e.g. FAQs already prepared)Implementation on an NLU platformTest on a panel of users (internal and/or external)Refinement of the User Experience through user feedbackObviously all this is very complex and is usually carried out by developers specialized in the sector, but with a little time and, above all, a lot of perseverance you can manage to create a chatbot starting from the documentation and tools provided by the most advanced NLU platform (according to me): DialogFlow.



Example of what Google Diagflow doesIn fact, the creature from Google, thanks Special Data  to a convenient web interface, is able to provide us with all the tools to create a correct flow of interaction with the customer.Furthermore, the convenient Training function allows you to refine the agent's skills thanks to the supervision of a human.Note the automatic integration with the Google Assistant: by clicking a simple button we will have at our disposal all the power of text recognition and natural language of the new Mountain View creature.The advantages of Virtual Assistants through chatbotsThere are thousands of web pages that already explain in detail the advantages of this approach to customer care so we will simply list a few:Scalability : is our customer care office no longer able to respond to all user requests? The answer is not to hire more staff but to automate the process of analyzing demand as much as possible and responding accordingly.Efficiency : stop exhausting email exchanges; the customer no longer has to wait until nine in the morning to get a response, the service is available 24/7Modularity : once an agent has been created, it will be enough to implement a few more lines of code to satisfy new needs, generated for example by a new product that you want to launch.Multilingual : Up to 25 languages ​​available.





Marketing automation: once the user has chatted on our Facebook page we are able to send targeted promotions to users based on their requests.Integration with Messenger and other chatapps: whoever chats with our virtual assistant has become our Messenger contact (or Facebook, if you prefer) and we can interact with him with all the tools that Facebook makes available to us (GPS position detection, sharing of data and photos, voice messages, Messenger campaigns, etc). Let's not forget that the same can apply to Telegram, Skype and many other messaging platforms.This was just a quick digression on the topic of Artificial Intelligence and its concrete applications. There would be much more to explore and study. If you have any doubts or would like more information on a topic that interests you, leave your comment, I am at your disposal.