Bicycle AI

Enterprise tool that improves KPIs for support and sales teams by automating responses and works across all helpdesks. (YC W’17)

Adit Gupta
Jan 5, 2018 · Unlisted

In the summer of 2017, I helped the young team at BicycleAI in their effort to make support teams super productive. The brief was very open-ended and I was tasked with determining the right product for this enterprise solution.

The only constraint was to build a design strategy that fit within the current workflow of the agents. I had to work with their in-house support and product teams to analyse this workflow and build a point of view.

We identified that a Chrome extension was the best way to work across every helpdesk and integrate into every agent’s workflow seamlessly. Here are a few problems I helped solve via design in the course of learning this business.

#1 : Improve productivity

Since this was a core pitch we made to our customers we had to focus all our attention toward this problem. Here are some tenets that helped us reach the solution after multiple iterations.

  • Reduce effort and errors in conversations with customers via automated responses
  • Automate repetitive workflows like collecting CRM data, logging bugs, capturing leads etc
  • Onboard teams with the least friction to collect data for our models
  • Easily triggered and dismissed based on our confidence

A contextual tooltip interface

Our tooltip style interface for agents

In order to work on all popular help-desk softwares our interface needed to adapt everywhere. A tooltip-based interface worked perfectly for this because of the following reasons.

  1. Our ML suggestions could appear contextually right next to the agent’s workspace
  2. We could choose to invoke it based on our confidence, thus becoming less obtrusive.
  3. It easily scaled to various agent workflows like repeated workflows, searching, viewing popular/recent suggestions etc.

Analytics and team management design for team managers on Bicycle AI

#2 : Tracking and team management

Another reason customers needed a product like Bicycle was to eliminate manual quality checks on agents responses. These involved time-consuming and primitive random sampling of conversations by agents. Bicycle’s dashboard promised to eliminate this by —

  • Collecting data about agent efficiency like ticket closure time, open conversations, response rates etc; crucial for support teams to track
  • Showcase ML enabled metrics that would differentiate Bicycle like topics or issues discussed, grammatical errors, tone, customer sentiment etc.
  • Showcase and bubble up anomalies in ticket closure by particular agents or overall performance of the team.

Colour palette for visualisation

We decided to use 2 palettes that complemented our brand blue colour for data visualisations. I chose these palettes based on extensive research around colours that are accessible and clearly distinguishable for colour-blind users as well.

#1 Primary palette — Numbers on the left indicate the amount of colours needed for the plot
#2 Secondary palette — Numbers on the left indicate the amount of colours needed for the plot

The following image articulates what a typical dashboard looked like for any team manager on Bicycle AI.


Adit Gupta

Written by

Product Design @Flipkart | Previously @Housing

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade