Ways we used AI to automate largest call center with 500k calls per month using Voice Chatbot / Voicebot
Our team was approached by largest and one of the fastest growing pizza chain in Eastern Europe with 700 franchisee stores in 12 countries and 0.5 bln USD revenue for 2020. Due to fast growth, they wanted automate their contact center in order to scale faster and keep customer satisfaction at the highest level. For that purpose we decided to implement AI voice chatbot support which could answer calls on a specific topics and redirect most important calls (complaints) to a specific departments.
Challenges with Existing Solution
Due to COVID-19 most of the business of the company (almost 70% of total orders) moved to pizza delivery. From one side that helped smaller stores to cover larger area without need of increasing in-store staff and restaurant size. From other side it brought new set of challenges, in particular:
How to manage all deliveries & customer requests;
How to find information about specific order and update customer on status;
How to unload contact center, without losing the quality of the service.
Solution.
Like in most of our AI/NLP projects we decided to select specific isolated area we want to focus on and work on its development. At first we just launched the bot for customers who placed order (as it was easier to track and predict what information is required). The customer was satisfied with the service and results, so we decided to expand the range of requests that the bot could handle. Whole process looked like this:
Launched chatbot in pilot mode (only for customers who placed order);
Collected first stats of its work;
Launched chatbot in a “production” mode;
Analyzed 1,000+ most popular requests;
Adjusted existing dialogue scripts;
Launched bot to accept calls for all customers.
Results.
As a result of project our customer could process up to 0.5 mln calls per month using our chatbot. With such huge number of calls we managed to reach up to 85% accuracy in request recognition. Estimated saving from robot implementation is expected to be 250k USD within first year and up to 1 mln USD for 2nd and 3rd year of operation.
As for the future plans, we plan to do next improvements:
Train bot to immediately recognise complaint and switch to dedicated “complaint team”;
Integrate courier location and ETA;
Further expand list of topics the bot can cover.