One of Poste Italiane's main touchpoints through which it could offer assistance to its customers was the virtual assistant 'Poste'. This powerful tool required considerable human effort from Poste Italiane in order to be maintained: the learning process on the phrases not understood by the virtual assistant was completely manual and many of these were not even analysed due to the huge amount of data coming from different channels on which the chatbot was installed.
A further complication was the fact that the activity was not scalable: if Poste Italiane had wanted to put the virtual assistant on new channels, this would have led to an exponential increase in the number of misunderstood sentences to be analysed manually and consequently a huge expenditure of man-hours as well as costs.
The most interesting projects in the field of Natural Language Processing are those with a large amount of content and a lot of interaction with end customers: Poste Italiane was a perfect case for us because of these characteristics. An abundance of data is usually positive in the world of artificial intelligence, but only if it is structured in the right way and monitored efficiently. This project had precisely the objective of working on these two aspects, creating an incredible mixture of Ai and people by exploiting what in jargon is called 'human-in-the-loop'.
Among Poste Italiane's values, customer satisfaction and innovation are two cornerstones for offering an effective and efficient customer service. For this reason, Poste Italiane chose to work with Indigo.ai to ensure that its virtual assistant "Poste" would achieve the best possible performance, automating the learning process through the use of Indigo.ai's proprietary platform while reducing the human effort of managing and maintaining the chatbot.
The project was completely managed through Indigo.ai's proprietary platform and consisted of three phases: organising the bot's knowledge base, improving the self-learning process and analysing the service quality comments left by customers at the end of the conversation. During the first phase, Indigo.ai managed through algorithms to optimally restructure the knowledge base (the brain) of the virtual assistant, leading to a 10% increase in the understanding of customer requests. In the second phase, previously incomprehensible requests were automatically qualified and clustered, resulting in a tool that reduced human effort on the training and maintenance process by 70%. Finally, in the third phase, customer feedback on the quality of the service was analysed, giving the Poste Italiane team further insights into how they could optimise the knowledge of the virtual assistant in order to satisfy their customers' needs as much as possible.
Increased understanding of customer requests by the bot
Reduction of the Full-Time Equivalent (FTE) on bot maintenance activities
Happiness rate of operators for chatbot maintenance
Experience the power of Conversational AI and unlock the hidden value in your data. We walk you every step of the way to the adoption of AI in your company.