Problem

Schedule managers were responsible for creating schedules that met complex operational requirements, but the existing experience required significant manual effort. Many of the rules that shaped a schedule lived in managers' heads or were repeatedly configured by hand. This made scheduling time-consuming and introduced opportunities for errors and inconsistencies.

The challenge was bigger than simply making the scheduling interface easier to use. We needed to create a tool with built in automation

How might we help schedule managers create accurate, compliant schedules with less manual work while still giving them the control and flexibility they need?

Goal

• Completely redesign the way schedule managers complete task through Oracle Agentic Application that automates repetitive staffing tasks—such as scheduling, balancing, and publishing

• Surfacing decision-ready issues in priority order

• provide guided recommendations that enable Schedule Managers to act quickly and confidently.

Challenge

My work included designing experiences for Shift Libraries, Schedule Generation Profiles, Work Patterns, and automated schedule generation, creating a connected system that allowed managers to move from manually building schedules toward a more streamlined and automated workflow.

Customer story

Meet Maya Devlin who is a schedule manager at an in-patient hospital for 5 different department locations.

Today, much of the work happens manually. They know the team's staffing requirements, recurring work patterns, and the rules they need to follow—but translating all of that knowledge into a schedule takes time.

Design process

I was given a list of metrics from my product manager as part of the requirement for this command center. Based on those metrics, I grouped and categorize similar metrics and came up with 3 main categories that would turn into agents: Schedule Coverage, Violations, and Request

What agents and what are they responsible for?

I mapped the staffing process to identify where an agent could reduce repetitive work and where human judgment should remain essential. This included what each agents metrics, insights, recommendations and actions associated.

Agent responsibilities

• Monitor staffing coverage

• Identify emerging gaps

• Check eligibility and availability

• Surface scheduling and compliance violations

• Recommend workers

• Identify opportunities to rebalance staff

• Prepare actions for manager review

User Flow & Wireframes

Next, I created low-fidelity user flows to plan out how these users flows would look like using our agentic app design system.

Making complexity scannable

Based on customer research, schedule managers want MORE context and transparency to understand how recommendations are made and sources to validate these decisions made by the agent. I used progressive disclosure to prevent the command center from becoming another overwhelming staffing tool.

The primary view answers three questions quickly:

What is wrong?
Critical staffing risks and violations.

How urgent is it?
Time-to-shift and severity indicators.

What can I do?
There will initially be a phase where users may not trust the agents. In cases where we recommend a worker to assign an open shift to

What are my other options?

Instead of having the user to search for alternatives, we display clear content as to how the agent came up with the recommendation and what were the other alternative recommendations.

Control over schedule changes

I utilized the chat experience to showcase how a schedule manager can have more control over schedule changes. I don’t think this is the best experience especially if the prompt is not written correctly or if the user text input isn’t specific enough, the agent may spit out wrong information.

Result

We’ve eliminated the 42 present options and instead give users the flexibility to pick exactly the timespans they want. Users don’t have to have duplicates because they can have control on what they want to see. With the redesign for relative timespans, users can use intuitive and simple drop-down selection which forms a sentence that matches natural human language.

Success metrics

This project is currently being implemented, but I worked with the Product manager to determine what those success metrics look like. Using clickstream data, we want to track

Clickstream data tracking

• Add relative dates

• Add specific dates

• Selections for date measurements

Next steps 

As mentioned earlier, this project ended up being scoped into three projects. I’ve completed timespans for Relative dates, now my next steps is to tackle timespans for Specific dates. For the next part of this project, I will used our date picker component provided by the design system (Stencil) which includes single dates and date ranges. The challenge for me is to use the date picker, Stencil and keep a consistent behavior and experience for all the other date pickers. I had to make custom components to full-fill the other date formats; Weeks, Months, Quarters, and Years.

Click through prototype