udligon_bot

Inspiration

Working on CRM we figured that Sales folks waste lot of time going through same emails again and again to find answers. We wanted something that can improve their productivity drastically.

What it does

It can parse email conversations and answer questions hidden in them. This can be extended to notes as well.

How we built it

We used Bidirectional attention flow model to build core ML model. API's were created on top of them to fetch and answer questions on email.

Challenges we ran into

Building an app for freshworks for the first time we faced challenges around available APIs and limitations on calling external APIs

Accomplishments that we're proud of

Building something useful in the first attempt.

What we learned

Building Freshworks apps is not difficult !!

What's next for udligon_bot

1. This can help auto fill many fields like product name, territory, etc. This information is usually hidden in mails and sales rep are lazy to fill.
2. Make the model lighter and faster
3. Integrate with freshcaller so that it can answer question based call log
4. Ability to answer across leads/contacts
5. Maintain privacy of sales rep. Answers should be restricted for a email/notes owner.

Built With

• allennlp
• python