Final project outputs include branding, UX & UI for multiple apps, and a full design system.
Form Field Structure
Users spent most of their time in cumbersome Windows 95 style legacy databases that often had unique and confusing restrictions or functionality that were added ad-hoc over time. The experience was ugly, but more importantly inconsistent and offered poor feedback.
Our team focused on building forms, tables, and smart searches from the ground up to cover the varied functionality across systems in a consistent user experience that was ADA compliant.
AI Predicted Data
A core deliverable across applications were several AI-driven data models that would aid users decision making and allow employees to spend less time on routine cases and more time on challenging cases.
Working with users and data scientists over several sprints, we refined the best UX approach to displaying AI selections that still empowered users to override the data model using their expertise.
Unified Item Search
Despite various systems depending on reliable equipment data to track every item at the power plant, they had no way of reliably searching or ID’ing the correct item.
Aside from unifiying various databases to centralize the data, Equipment Search (and later search for parts, inventory, and more) required a smarter and unified modal system that could live across applications and devices.
Just as important as building the foundation for the applications was documenting the styles, components, and usage of the core elements. Our team constantly updated our style guides and explainers in Invision and Zeplin after every sprint, and often released new documentation with feature launch meetings that allowed executives and implementers across departments to give input and gain insights.
Our team spent multiple sprints focused on wireframes for hundreds of screens, which changed and improved on weeks of user testing and SME input. Eventually these wireframes boiled down to four core app journeys primed for high fidelity design.
By the end the Work ID phase (26 two-week sprints), we had created more than 20 user journeys, over 100 unique screens across four device breakpoints (Desktop, Legacy Desktop, Tablet, Mobile).
Screening Work Queue
Our team worked closely with users to reimagine a daily meeting (featuring a 100+ page printed document of hand entered tasks) into a digital work queue. Multiple personas ended up benefitting from this queue view, and as we tested the app with more users additional use cases found their way into the app.
On the backend, AI models analyzing the large historical record of tasks allowed us to suggest changes and adjustments to tickets.
AI Feedback
It was vital for screeners — some of the most highly trained and senior employees on a power plant floor — to be able to not only receive AI suggestions but also overwrite them.
Throughout the engagement, our team refined how AI suggestions we're displayed and how easy they were for users to change. When the expert users made changes to the AI suggestions, the app captured feedback from users in real time to feed back into the model for improvement.
Reducing Redundant Work
In the fog of user interviews, our team found a consistent time waster was duplicate work. When an item at the power plant has an issue, there were often multiple individual tickets generated from different employees or departments. In the old manual process, sorting through these duplicates was messy and time consuming.
In the Task Screening Queue, our team created a way to bundle duplicate tasks and take the best information from each ticket. In addition to AI-driven bundling, users could create their own — with that data fed right back into the AI models to improve it's ability to identify duplicates.
In the scheduling phase our team produced two core tabs for separate persona’s, with multiple user journeys for each, across over 30 unique screens.
