01
Understand
Learn how Schedule Managers balance schedules today, where they experience friction, and what they need to make confident staffing decisions.
Oracle · Product Design · AI-powered scheduling
Reimagining workforce scheduling with AI.
I designed an agentic scheduling experience that helps Schedule Managers automate repetitive work, quickly identify issues, and focus on the decisions that need human judgment.
HealthcareManufacturingRetailUX strategy · Research · Product design
The opportunity
As the product strategy shifted toward AI, I researched user needs, explored three distinct design directions, and presented them to the VP of Product for Human Capital Management. The work helped the team align on an AI-first experience that prioritizes issues and pairs them with contextual recommendations.
10–30 min
The project’s envisioned daily scheduling workflow, compared with weeks of planning and 10+ hours of balancing and rebalancing.
01 / Overview
Designing through ambiguity
This project took place during a period of rapid change as the business shifted its product strategy toward AI-powered experiences. Scope and requirements changed multiple times, creating a challenge that called for research, strategic exploration, and adaptability.
02
Explore
Evaluate traditional, hybrid, and agentic directions against customer needs, automation potential, development effort, and product strategy.
03
Deliver
Design a smart scheduling experience that prioritizes issues and offers contextual recommendations, helping managers act faster and with greater confidence.
02 / Background
The product vision changed mid-project
The work evolved from improving one stage of the scheduling cycle to exploring automation across the full scheduling lifecycle.
Phase 1
From traditional scheduling to hybrid AI
The original scope focused on the balancing phase: adding or removing shifts based on staffing needs, reallocating resources through transfers or float pools, and considering external agencies before schedules were published.
The initial strategy was to build on Oracle’s Redwood framework and introduce AI where it could provide meaningful value. Questions remained around how customers would use AI credits, how the experience would be priced, and how much customers would be willing to pay.
Phase 2
Automating the scheduling lifecycle
The scope expanded to schedule creation, balancing, publishing, and ongoing management with exemptions. The strategy shifted away from a hybrid approach toward an AI-first, agentic experience through Oracle’s Agentic Application.
Because the product and its capabilities were evolving alongside the project, I had to adapt the design approach while staying focused on users’ needs.
03 / The problem
Staffing decisions were fragmented and reactive
Problem statement
Traditional staffing and scheduling rely on manual coordination, disconnected tools, and reactive decisions. Managers must quickly identify coverage gaps, confirm employee eligibility and availability, and rebalance staff across five departments as patient demand changes. The process is time-consuming, difficult to scale, and can lead to inefficiencies and inconsistent decisions.
Primary user
Schedule Managers coordinating staffing across multiple department locations.
Core need
A unified view of staffing conditions and clear, trustworthy guidance on where resources can be reallocated.
04 / Research
What I set out to learn
I led three research sessions through an internal Design Critique group. Participants were Oracle employees with inpatient hospital experience or profiles similar to our primary users.
01
What do Schedule Managers need today?
Validate concepts based on customer-requested needs and existing Redwood design patterns.
02
What could they need tomorrow?
Explore an agentic scheduling experience that customers had not explicitly asked for, but that could address deeper pain points.
Research goals
05 / Key findings
Three insights shifted the conversation
01
A unified view enables confident decisions
Participants responded strongly to a cross-department schedule view they called a “House View.” It helped reveal where resources could be reallocated.
“I need a unified view of all the schedules so that I can efficiently reallocate nurses between departments where there is a need.”
Design implication
Showing open shifts alone was not enough. Managers also needed shift, job role, competencies, float usage, acceptance percentage, call-outs, and other staffing indicators.
02
Users were ready for AI
We expected some resistance to a new AI-driven workflow. Instead, participants were receptive—and often excited—about the possibility of reducing repetitive scheduling work.
“I’m tired of balancing…”
Design implication
The opportunity was not just to make balancing easier, but to reduce how much manual balancing managers needed to do.
03
The Command Center felt familiar
Schedule Managers already understood a centralized hub for shift coverage, labor allocation, open shifts, and float availability.
Design implication
Position AI as an intelligent layer in a familiar, centralized experience instead of introducing an entirely unfamiliar way of working.
Original question
How might we help Schedule Managers balance schedules more efficiently?
Research-led question
How might we automate the scheduling lifecycle so managers spend less time managing schedules and more time making confident staffing decisions?
06 / Design exploration
Three directions, one strategic decision
I explored three possible directions, balancing customer control, automation, development effort, and long-term product strategy.
DIRECTION 01
Buildable + Enhancements
Improve the existing Redwood experience instead of fundamentally changing the workflow.
Customer impact
Preserve visibility and control while reducing the need to jump between schedules and streamlining existing tasks.
Research feedback was positive, but the full experience could require more than two years of development and risk falling behind an increasingly AI-driven market.
DIRECTION 02
Hybrid Experience
Combine familiar Redwood workflows with AI recommendations at moments where they can provide the most value.
Customer impact
Keep managers involved in decisions while helping them identify potential assignments faster.
A lower-risk transition to AI, but it would require maintaining traditional experiences alongside the agentic model. This direction was not directly tested with customers.
DIRECTION 03
Agentic Application
Automate repetitive scheduling work and surface the issues that need human attention.
Customer impact
Shift from managing every task to overseeing exceptions and making important decisions, with human judgment kept in the loop.
V1 also needed to preserve critical familiar workflows, including assigning shifts and responding to Open, Claim, Drop, and Swap requests.
| Direction | Customer control | Automation | Strategic alignment |
|---|---|---|---|
| Buildable + Enhancements | High | Low | Low |
| Hybrid Experience | High | Medium | Medium |
| Agentic Application | High, human-in-the-loop | High | High |
Research suggested customers were open to AI taking on more scheduling work. Combined with the evolving product strategy and competitive landscape, this supported moving toward the Agentic Application direction.
07 / Design process
Turning complex requirements into a clearer experience
Once product requirements were updated, I organized required metrics into common themes to define what information agents should provide and how it could be made digestible.
1
Define the information
Gather the required metrics and group them into themes to establish the content needed for the experience.
2
Explore the presentation
Use rapid prototypes to explore ways to present a large volume of information clearly and actionably.
3
Adapt as the product evolved
Refine the approach as scope, requirements, product strategy, technical possibilities, and the design system changed.
The power of vibe coding
Using AI to accelerate exploration
I used ChatGPT with a skill that turned my product manager’s Confluence requirements into a working prototype. With so many metrics to consider, this helped me explore more digestible ways to present information and test concepts against the Agentic design system.
It was not always accurate—some components did not match the design system—but it was useful for generating inspiration and communicating an idea early.
08 / Challenges
Learning fast while navigating constant change
Designing an agentic scheduling experience meant building new expertise while keeping the team aligned through a rapidly evolving product direction.
CHALLENGE 01
What exactly is an agent?
Agentic AI was new territory for me. Alongside my design work, I needed to quickly understand what an agent is, what it can do, and how it could create meaningful value for our customers.
How I approached it
I enrolled in classes and online training through our internal learning resources, building the knowledge I needed to make more informed design decisions as the project progressed.
CHALLENGE 02
Designing through a moving target
Project scope, requirements, product strategy, and even the design system continued to evolve. This made it challenging to maintain momentum while designing toward a shared direction.
How I approached it
When repeated scope changes began affecting planning, I involved my manager to help coordinate a meeting with the product team and realign on scope and timeline. I also partnered closely with the Agentic Application designers to validate the end-to-end flow and apply the right interaction patterns.
09 / Final solution
From managing schedules to managing decisions
The experience is designed to help managers see the issues that matter, understand the context, and take action with AI-supported recommendations.
Scheduling command center · concept illustration
Staffing overviewNeeds attention
Coverage82%
Open shifts14
Suggested moves6
Priority issues
Cardiology
ReviewCoverage gap · best-fit workers identified
Emergency
SuggestedShift needs coverage · 3 eligible workers
Medical / Surgical
SuggestedFloat availability detected
Illustrative concept only. Replace with approved product screenshots before publishing.
The agent handles the repetitive work. The Schedule Manager handles the decisions.
Agents can take on repetitive scheduling work and provide best-fit, availability-aware recommendations. Schedule Managers retain oversight and step in when human judgment is needed.
The broader vision also connects workforce, talent, skills, payroll, and costing insights so managers can adapt to changes and fill open shifts with more context.
ADD PRODUCT VISUALHouse View, resource allocation, or agent recommendation flow
10 / Impact
Giving managers time back
AI was changing both how I worked as a designer and how customers could work. The opportunity was to rethink the workflow itself—not simply add AI to an existing process.
10–30 min
Daily scheduling work
The project’s stated outcome frames work that traditionally took weeks of planning and 10+ hours of balancing and rebalancing as a short daily workflow.
Best-fit
AI-powered recommendations
Help managers identify suitable, available workers to cover shifts, reducing the burden of finding the right assignment.
Connected
AI insights
Help managers respond to change with connected HCM, talent, skill, payroll, and costing insights, guided by an intelligent assistant.
Before publishing, confirm whether the 10–30 minute figure is a measured result, target, or intended product outcome and label it accordingly.
11 / Reflection
Designing for change is part of designing the product
AI did more than change the tools I used as a designer. It challenged me to rethink how our customers could work—and where human attention creates the most value.
Stay grounded in user needs
When strategy and technology shift quickly, research helps separate a changing solution from the underlying problem worth solving.
Use AI to explore, not replace judgment
Rapid prototyping helped make complex requirements tangible. The designer’s role remained to assess the output, align it with the design system, and make intentional decisions.