Automated E-mail Response System For Query Resolution

Automatic email reply with appropriate product specification document

Table of contents
Shilpa Ramaswamy

The Client & the Challenge

One of the largest multinational chemical companies with 50,000 plus employees and present in more than 50 countries was looking for an automated email query resolution system.

Industry Overview


As the chemical industry moves into 2022, strong demand for both commodity and specialty chemicals should keep prices robust throughout the year. The industry should also experience increased capital expenditure as leading industry players focus on building capacity and expanding into growing end markets through both organic and inorganic routes. One of the critical areas of focus for most chemical companies will likely be sustainability and decarbonization. Many companies are expected to increase investment in research and development (R&D) capabilities and leverage advances in decarbonization and recycling technologies to lower their and their customers’ carbon footprint, as well as reduce plastic waste.

Business Challenge

Customer Experience

Automated customer service and support are booming – and for good reason. The client wanted a solution to establish an automated email query resolution system to provide tailored experiences for potential customers because they’re based on data about their behavior or actions eliminating the need to manually send out individual emails that would save time, energy, and resources. The client deals in a plethora of products ranging from prescription medicines to vaccines to healthcare products, and more, and customers might require additional informational documentation to use such products. The client wanted us to send an appropriate product specification document from their website if the email query had requested the same.

The client wanted to establish an automated email query resolution system that provided tailored experiences for potential customers based on data about their behavior or actions.


We used a blend of vision AI and Deep Learning to solve the customer's challenge. Here is a breakdown of the steps we used:

Step 1: Understand the email intent and extract relevant information

One of the primary challenges was to understand the email intent if the email query was to get the appropriate product specification document. This was automated by training a binary classification model by using two different types of email texts; email text for the product specification document requirements and email text for any other purpose. After having figured out by using the classification model that if the email query is intended to get the product specification document, the relevant information is extracted from a customer email query such as product name, invoice id, and other vital specifications. For any other type of email, the query was resolved using manual intervention.

Step 2: Trained the Name-Entity-Recognition model

NER (Name Entity recognition) model was trained to automatically annotate entities in the email such as product id, catalog id, reference-id, and more to clearly identify all the required attributes to fetch the relevant specification document, and the entities were saved in the database.

Step 3: Extracted the product specification document from the database

Product id, reference Id, and similar details were used to extract the product specification document from the backend. The required document was then attached to the email query response and was settled by an automatic email reply.

Impact Delivered

  • 55% of the email responses were automated.
  • Provided specifications for 13000 products.

Top Benefits

  • Around-the-clock support
  • Lower operational costs
  • Faster response times

The Akaike Edge

Inbuilt libraries, DL models with transfer learning capabilities