Product: Conversational AI for Hotels

The Product


The PolyAI partnerships team launched a repeatable voice application for hotel front desk services.

My Role

Summer Internship during the Cambridge MBA. The scope of work was to identify the bottlenecks for scaling our product operations.

The Team

3 machine learning engineer 2 back end engineer 1 team lead 1 dialogue designer 1 product manager

Project 1: Standardizing Process

The problem - App development was crafted around personalized voice applications. This new use case required PolyAI to deliver faster and thus branching out a different development process.

The discovery - Through analyzing the call and intent data of 7 live versions of the voice application, I was able to identify 24 FAQs out of 150, that addressed 90% of the call volume.

The result - This insight enabled us to refactor bottlenecks in our development process reducing time-to-go-live from 3 weeks to 3 days.

Project 2: Defining key operational metrics

The problem - PolyAI had 7 different live versions for the same use case, and had 200 customers in the implementation pipeline. The scope of work was - how do we define what quality is? and quantify it going forward.

The discovery - I evaluated the performance of the distinct versions according to the typical 3 key metrics of call cneters: effectivity, efficiency and customer satisfaction. Then, compared the versions amongst each other to establish a baseline value and confidence intervals.

The result - The dashboard highlighted voice agents behaving out of the norm, enabling the partnerships team to intervene when needed. Six months later the programme accumulated 1.1 million customer calls with 84% serviced without a human and 80% positive customer ratings.