Powered by Office 365 Graph
As a leader in today's business landscape, we often get inundated with emails from our team, employees and even customers around how they feel about a particular meeting, initiative, idea or culture at your company. These emails can be time consuming. Additionally, it's counterproductive to respond to each message (and it's emotional gravity) without getting carried away in biases and sentiments. We surveyed our friends and colleagues around how much time they spent with feedback-driven email communication - and the pain-point was very real! This is where we got inspired to build Pulse.
- Specific Feedback: Gather feedback and sentiments on what is important to you. #townhall, #christmasparty, #benefits, #strategy
- Individual Feedback: Gather feedback on your performance and goals. #leadership, #communication, #teamwork
- General Feedback & Comments: Gather feedback and/or comments on any topic that is important to you. #hoursofoperation, #customerservice, #dresscode
- Idea Corralling: Gather and get a birds eye 'feel' of new ideas. #iPhone8, #Safety, #Revenue, #CustomerPortal
- Climate Survey: Gather sentiments about the organization, leadership, your team, and your role. #culture, #collaboration, #values, #change.
What it does
Pulse is an AI-powered app that feeds off of your Office 365 Graph (Outlook Mail) to analyze emotions and sentiments (and more in future AI models) of messages that hit your inbox. Here are some highlights of how the app works:
The #HashTag Habit
By leveraging the popular #hashtag habit from social networks, leaders ask their audiences to include key topics and organizational keywords. This flags our AI to "listen" and parse emotions from the appropriate messages!
Search by Topic
You can always search by a specific topic (no #hashtag needed!) when you need to analyze a particular topic/keyword from your inbox.
Responding to difficult (and emotionally charged) emails as well as crafty tone-savvy messages can now be done with our Compose Assistant right from the dashboard.
How we built it
We built it using NodeJS, Azure, Microsoft Graph APIs and custom AI models we built using Natural and IBM Watson technologies.
Challenges we ran into & what we learnt
- Handling auto-renew of the Graph Subscription webhook endpoints without duplication, while adhering to the user's authentication token lifecycle.
- Integrating AI at the right level - not scanning an entire email, detecting signature text (noise) as well as building a model that can scale beyond the emotions currently classified via our AI.
Accomplishments that we're proud of
- Climate surveys, feedback tools and similar leadership technologies take a ton of time (and money). Additionally, email is a ubiquitous medium in the enterprise. By hacking into the email flow using the Graph APIs, we're able to focus on solving for productivity v/s building unnecessary UI elements and experiences.
- The architecture of the solution and the AI models will allow us to scale the app in the future with features like notifications and "emotional alerts" based on certain thresholds in a leader's inbox and additional AI classifiers.
What's next for Pulse
- Kick-off the business model: Based on our initial survey of Pulse's IQ-as-a-Service, we're currently enrolling enterprise 'paid betas'.
- Building the Email Dashboard: Taking our AI beyond emotion/sentiment analysis will allow us to detect tasks, automation and more in a leader-focused dashboard to their inbox.
- Voice UI: We're also working on Cortana and Alexa skills to allow leaders to ask for a daily summary of their email's "pulse" via voice!
- Bias Reduction: Certain individuals have an inclination to be sarcastic. By allowing the leader to flag such emails / conversations, our AI will be able to learn individual 'sentiment patterns' and allow more accurate sentiment analysis as well as email crafting
- Native Add-ins for Office 365.