Project Description
E-MAIL ROUTING
AEVITAE
BRIEF
Aevitae’s customer service receives hundreds, if not thousands of e-mails per day. Up until recently, every new e-mail was read by an employee and manually forwarded to the relevant department. Not only was this process time-consuming, but it was also costly. The question was: can we tackle this problem smarter and more efficiently? Within six weeks, our Data Science team developed a machine learning model that automatically forwards 90% of the incoming e-mails to the right department. The AI tool can spare up to 2 FTEs per year. During this project, we assisted Aevitae with:
Developing an AI tool that automatically forwards e-mails to the right department
- Creating a Power BI dashboard that displays the numbers of forwarded e-mails and how often this process is performed correctly
![Aevitae-E-mail-Routing-Dashboard](https://usercontent.one/wp/mediaan.com/wp-content/uploads/2020/09/Aevitae-E-mail-Routing-Dashboard.png?media=1714647656)
![icon_concept](https://usercontent.one/wp/mediaan.com/wp-content/uploads/2019/02/icon_concept.png?media=1714647656)
THE CONCEPT
Every day, up to thousands of incoming e-mails were manually forwarded to the right department. An AI tool can automate this repetitive task by learning from the historical data. For example, by looking at the e-mails that Aevitae has filtered and categorized over the past year, the AI tool can use machine learning techniques to predict which department should pick up the e-mails.
![icon_challenge](https://usercontent.one/wp/mediaan.com/wp-content/uploads/2019/02/icon_challenge.png?media=1714647656)
CUSTOMER CHALLENGE
Promptly answering all of the customers’ questions was challenging, particularly during busy periods. Because all e-mails arrive in one main folder, it requires the customer service employee to drag each e-mail to the correct folder manually. This way-of-working was time-consuming and created delays in the process.
![NLP-AI-robot](https://usercontent.one/wp/mediaan.com/wp-content/uploads/2020/09/NLP-AI-robot.png?media=1714647656)
MEDIAAN IN ACTION
With our expertise in Natural Language Processing (NLP), our team developed an AI tool. The following techniques and technologies were used:
Python
- pandas
- scikit-learn
TFIDF Transformer
- Vectorizer
Vectorizer
Support Vector Machine
- exchangelib
- Microsoft PSQL Server
Microsoft Power BI
NEW INSIGHT
Our team also designed a Power BI dashboard that showcases the numbers of forwarded e-mails. It also displays how often this process is performed correctly and when an employee had to correct it manually. The new ability to track the number of e-mails each department receives and how it evolves over time provides Aevitae with brand new insights into their customers.
![Correct-Email-Fowarding](https://usercontent.one/wp/mediaan.com/wp-content/uploads/2020/09/Correct-Email-Fowarding.png?media=1714647656)
![icon_result](https://usercontent.one/wp/mediaan.com/wp-content/uploads/2019/02/icon_result.png?media=1714647656)
RESULTS
Within six weeks, our Data Science team designed, built, and trained a machine learning model that automatically forwards 90% of the incoming e-mails to the right department. The AI tool can spare up to 2 FTEs per year. Two weeks after the implementation, 77% of the incoming e-mails were forwarded to the correct folder. The customer service employees manually corrected the remaining. Due to the self-learning component of the model, the percentage will only keep getting better over time. One year after going live, the tool forwarded 100% of the e-mails, of which 88% was forwarded correctly.