Workforce Planning | Andrew Corey

Predictive Scheduling

The Predictive Scheduling module enriches existing scheduling tools with demand based forecasts and smart balancing helping nurse leaders easily create accurate optimized schedules.

Company

Product

Role
Duration
Platform

Company

Product

Role
Duration
Platform

Aya Healthcare

Predictive Scheduling
Senior Product Designer (IC)
2024 - Present
Enterprise Web (B2B Saas)

Aya Healthcare

Predictive Scheduling
Senior Product Designer (IC)
2024 - Present
Enterprise Web (B2B Saas)

Impact

Enabled organizations to schedule staff proactively rather than reactively
Optimized shift coverage across multiple departments, reducing last-minute shortages
Contributed to smoother operational planning as part of Aya’s Workforce AI suite

Problem

Healthcare scheduling is complex:

Multiple roles, shift types, and locations
Manual scheduling leads to last-minute coverage gaps
Reactive decision-making increases operational risk and staff burnout

The challenge:
Design a predictive scheduling experience that anticipates staffing needs, surfaces conflicts early, and allows flexible human oversight — without overwhelming the user with data.

Healthcare scheduling is complex:

Multiple roles, shift types, and locations
Manual scheduling leads to last-minute coverage gaps
Reactive decision-making increases operational risk and staff burnout

My Role & Scope

I led the product design for Predictive Scheduling, partnering with:

Product Management
Engineering
Data Science
Operational leaders
Executive stakeholders

Responsibilities:

Defining the end-to-end scheduling experience
Translating predictive algorithms into actionable scheduling recommendations
Aligning workflows across operational and executive users
Rapid prototyping for iterative testing and stakeholder alignment

I led the product design for Predictive Scheduling, partnering with:

Product Management
Engineering
Data Science
Operational leaders
Executive stakeholders

Responsibilities:

Defining the end-to-end scheduling experience
Translating predictive algorithms into actionable scheduling recommendations
Aligning workflows across operational and executive users
Rapid prototyping for iterative testing and stakeholder alignment

Users & Context

Primary users:

Staffing coordinators
Workforce planners
Department heads

Users needed to:

Quickly identify conflicts or gaps in staffing
Adjust schedules efficiently while considering predictive insights
Maintain transparency and fairness for staff

Key constraints:
Errors in scheduling could lead to understaffed shifts or operational disruption, so trust and interpretability were critical.

Primary users:

Staffing coordinators
Workforce planners
Department heads

Users needed to:

Quickly identify conflicts or gaps in staffing
Adjust schedules efficiently while considering predictive insights
Maintain transparency and fairness for staff

Key constraints:
Errors in scheduling could lead to understaffed shifts or operational disruption, so trust and interpretability were critical.

Key Constraints &

Design Challenges

Key Constraints & Design
Challenges

Prediction accuracy vs human judgment: Balancing AI recommendations with manager discretion
Data complexity: Many roles, shifts, and constraints must be visualized clearly
Workflow fit: Schedules needed to integrate with existing staff systems
Trust in automation: Users had to feel in control of AI-assisted schedules

These constraints informed every interaction, visualization, and workflow decision.

Design Approach

Predictive Scheduling is a system that moves teams from static schedules to AI-assisted, continuously optimized schedules — while preserving human control.

Scenario-based scheduling: Users can simulate changes (shift swaps, staffing adjustments) to see impact

Progressive disclosure: Only critical information is surfaced upfront, details available on demand
Contextual recommendations: AI predictions come with rationale and confidence indicators
Integration: Scheduling recommendations embedded directly into existing workflows to minimize friction

Close collaboration with Data Science ensured recommendations were accurate, actionable, and explainable.

Key Design Decisions

1. Make scheduling recommendations actionable

Recommended shifts presented with clear rationale

Visual indicators of risk, coverage, and priority
Users can accept, modify, or reject suggestions

1. Make scheduling recommendations actionable

Recommended shifts presented with clear rationale

Visual indicators of risk, coverage, and priority
Users can accept, modify, or reject suggestions

2. Scenario simulation and “what-if” analysis

Users can adjust staffing assumptions and see predicted outcomes

Allows proactive staffing decisions without waiting for crises

2. Scenario simulation and “what-if” analysis

Users can adjust staffing assumptions and see predicted outcomes

Allows proactive staffing decisions without waiting for crises

3. Highlight conflicts and gaps visually

Color-coded alerts for overlapping shifts or undercoverage

Visual hierarchy guides user attention to the most urgent issues

4. Progressive disclosure for trust

Advanced AI details (model confidence, reasoning) hidden behind expandable panels
Avoids overwhelming operational users while keeping transparency for curious managers

Predictive Scheduling System Overview

Predictive Scheduling

System Overview

A modular scheduling system supporting creation, optimization, and continuous adjustment
Designed Predictive Scheduling as a set of connected modules, each optimized for a different scheduling decision.

These three modules are a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

Schedule Template Overview

Schedule Template

Overview

Encoding scheduling intent and constraints upfront
Designed templates to capture real-world staffing needs before introducing optimization.

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

Encoding scheduling intent and constraints upfront
Designed templates to capture real-world staffing needs before introducing optimization.

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

Base Schedule Template Detail

Base Schedule

Template Detail

Constraining optimization to reflect operational reality
Ensured AI outputs respected labor rules, coverage requirements, and team preferences.

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

Constraining optimization to reflect operational reality
Ensured AI outputs respected labor rules, coverage requirements, and team preferences.

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

New Optimized Schedule Template

New Optimized

Schedule Template

AI-generated schedules as a starting point, not a final answer
Positioned optimization as a baseline that users could review, adjust, and approve.

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

AI-generated schedules as a starting point, not a final answer
Positioned optimization as a baseline that users could review, adjust, and approve.

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

Optimization Explanation / Metrics

Optimization Explanation

/ Metrics

Why the schedule was optimized this way

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

Why the schedule was optimized this way

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

Optimized Schedule

Completed Optimized Schedule

Completed Optimized

Schedule

Built trust through transparent optimization logic
Explained tradeoffs and improvements to help users evaluate AI recommendations.

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

Built trust through transparent optimization logic
Explained tradeoffs and improvements to help users evaluate AI recommendations.

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

Auto-Balance Dashboard

EDIT
EDIT

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

EDIT
EDIT

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

Auto-Balance in Action

Maintaining coverage as conditions change
Designed Auto-Balance to respond to real-time disruptions while minimizing unnecessary schedule churn.

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

Maintaining coverage as conditions change
Designed Auto-Balance to respond to real-time disruptions while minimizing unnecessary schedule churn.

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

Auto-Balance Control & Overrides

Auto-Balance

Control & Overrides

Preserving human control in automated workflows
Gave users explicit control and visibility to maintain trust and accountability.

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

Preserving human control in automated workflows
Gave users explicit control and visibility to maintain trust and accountability.

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

Completed Auto-Balanced Schedule

Completed Auto-Balanced

Schedule

EDIT
EDIT

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

EDIT
EDIT

Frame the three modules as a progression, not features:

Schedule Template → Define intent and constraints
New Optimized Schedule → Generate an AI-assisted baseline
Auto-Balance → Maintain coverage as conditions change

Solution

The final Predictive Scheduling experience included:

Proactive scheduling recommendations with confidence scores
Visual alerts for conflicts, gaps, and coverage risks
Scenario simulation for planning adjustments
Integrated workflows for operational and executive users

The final Predictive Scheduling experience included:

Proactive scheduling recommendations with confidence scores
Visual alerts for conflicts, gaps, and coverage risks
Scenario simulation for planning adjustments
Integrated workflows for operational and executive users

Product Experience Overview

Product Experience

Overview

Together, these modules enable teams to define scheduling intent, generate optimized schedules, and maintain coverage over time — transforming scheduling from a static artifact into a living, adaptive system.

Outcome & Impact

Operational teams reduced last-minute scheduling conflicts
Improved staff satisfaction and reduced burnout
Enabled proactive management of staffing resources
Contributed to enterprise adoption and demonstrated Aya’s AI value to clients

Reflection

Key takeaways:

Predictive tools must balance AI insight with human oversight
Visualization and hierarchy are critical in dense scheduling data
Scenario-based design encourages proactive behavior and increases adoption

If revisiting:

Explore personalized recommendations per department
Enhance predictive alerts with historical context
Expand executive reporting options for cross-team visibility

Let's work

together.

Let's work

together.