AI Platform
Designing an ecosystem for managing automation.
During my year-long engagement with this financial healthcare company, I was asked to tackle complex problems in support of a singular goal: simplify hospital billing through automation.
Overview
This project embraced a challenging goal of automating previously manual processes in hospital billing through a network of bots.
As the in-house technology team at this Healthcare Company began authoring a series of bots, it quickly became apparent that a unified solution was needed by the business for monitoring and auditing the reliability and effectiveness of the bots.
Business Goals
Create a system for tracking, auditing, troubleshooting, and reporting on the health of bots.
User Objective
Get on-time information from a single application that brings together many disjointed processes.
Technical Constraints
The UI needed to scale from multiple-display user stations to large television screens in war-rooms,
Creating a system to monitor bot health
The overall system was organized into several functional areas, each with their own set of business requirements and user types.
Rapid iteration through weekly design sessions
I met weekly with a team of business analysts, executive stakeholders, and technologists to ideate on different aspects of the experience. In the early stages we would whiteboard, sketch, and document the ideas that emerged.
Wireframing
For key screens, I created low-fidelity wireframes to drive definition of the intended functionality.
Defining UI Styles
A dark UI approach was selected to alleviate eye strain at workstations where users would engage with the experience for extended durations. The font was selected for its ability to display tabular data on dense screens.
The solution
The following screens represent the core functional areas that where explored. Several hundred screens where presented via prototypes to arrive at the end solution.
The treemap provides an overall view of the system of bots. Color coding is used to highlight bot anomalies with the largest impact to the business. Since it is anticipated to have some anomalies, the color coding was intentionally neutral to avoid inferring that there is an issue.
The bot details view contains an interactive radar chart - allowing users to inspect different criteria that make up the overall "health" of the bot.
Each spoke of the radar chart has unique criteria for what constitutes the normal threshold. Users are alerted to spokes that need attention.
Anomalies in bot performance can be quickly narrowed in upon, allowing users to inspect and report upon suspicious behavior.
A robot set of filters can be expanded to allow users to quickly refine and "slice" the data.
A built-in help-desk system for reporting on incidents impacting the health of bots was added in order to ensure instant escalation when an issue was identified.
Tracking incidents through completion allows business stakeholders the ability to quickly work on resolving issues as they arise.
Some issues required users to handle resolution behavior outside of the system, via a phone call, meeting, or using a IT software. Allowing for quick reporting on non-system actions made it easy to connect these actions.
Random sample audits are preformed on a daily basis to monitor and verify the accuracy of the bots. Since the system was brand new, the business demanded human oversight as a failsafe.
Auditing the actions of the auditors adds an extra layer of human oversight to the system. Additional table actions allowed users with more advanced settings to leverage the same UI with more tools available.
Auditors are presented with a snapshot of the task being automated. They are required to render a decision of how the task should be preformed in order to compare against the actions of the bot.
An overall view of the different bot configurations allows a instances of a single bot to preform a variety of different tasks.
Understanding the rules for each bot allows the end user to monitor if the bot is preforming as expected.
Advanced users are granted edit access to bot rules, allowing them to change the overall function of an individual bot.
The business operations view allows in-depth analysis of the performance of a bot over time. By filtering different criteria, the analyst can obtain a more holistic view of the bot.
Made for TV
Team “war rooms” where created, where large TVs would display vital system information to allow at-a-glance monitoring of the performance of the bots. Creating configurable UI panels allowed unique views to be compiled for the large screen displays. Additional views like the Kanban issue tracker allowed executive stakeholders a quick view of issue resolution.
A UI kit to support it all
In order to enable our front-end focused agile team, an exhaustive UI kit was created and evolved over the life-cycle of the project. The end result? A system that looked as smart as it functioned.
