Why is FreJun Moving to Deep Learning? A Step towards Happier Meetings

frejun deep learning system

We think you should know the background of FreJun to understand why we have decided to move into Deep Learning space for managing your meetings.

We will take you through the versions of FreJun over the years.

Version 1: The Naive “Typing In” Interface

We launched the first version of our product in September 2018 to collect the minutes of meetings and action items. The product was pretty basic. 

FreJun would send emails to all the meeting attendees that are on the calendar to set the agenda, collect the minutes, and actions. An aggregate email with all the details was sent to the user and attendees after the meetings. 

The problem with this version was the interface. Our users did not like typing interface as they have to stop during and after the meeting for collecting the minutes and action items.

As you can see, we just had the “Typing interface” without any additional features. It was easy to use but lacked some essential features. 

Version 2: The Voice Interface

After a couple of months of work, we deployed the voiced based command system. The user asks the inbuilt voice bot to do all the activities like setting the agenda, collecting the meeting minutes and action items

Even though the users liked the system, there was a concern. The primary complaint was that they didn’t want their meeting flow to get interrupted, and also there was a slight lag between the user’s commands and the system. Our users didn’t want to stop their train of thoughts during a serious discussion to give a command to the voice bot. 

Along with that, the users were feeling annoyed with the constant emails for setting up the agenda, collecting the minutes and action items. 

We also made a couple of design mistakes, which should have been avoided in the first place.

Ultimately we realized that this is not the user experience we want to deliver to our users. But this gave us a fantastic learning experience about user needs and expectations! 

Voice to Text interface

“Building the right product > Building the product right” – Srikrishnan Ganeshan

Version 3: Voice-based Deep Learning System

The third and the current version that’s under development is Voice-based Deep Learning System. The new product is simple to use with just three steps. 

1. Record your meetings

2. Open the link that you get for each meeting.

3. Export the summary and action items to different work tools like Jira, Asana, and SFDC, etc.

To provide the best user experience where the user has a minimum input and information loss, we have decided to build Deep Learning systems that extract the action items and summarizes the meeting automatically 

Leave a Reply

Your email address will not be published. Required fields are marked *


Enjoy this blog? Please spread the word :)