← Back to Resume
Workday VNDLY  ·  June 2026

Actions Engine — Interaction Patterns & Phased Rollout

🔒 NDA — Visuals password protected
Interaction Design Design Systems Human-AI Interaction Enterprise UX Agentic UX Product Strategy B2B
My Role
Lead UX Designer
Company
Workday VNDLY
Delivered
June 2026
Audience
Product & Engineering

Visuals are password protected

This project is covered by an NDA. Enter the password to view the prototype recording.

Incorrect password. Please try again.

Not every task is equal — but today our UI treats them that way

Within Workday VNDLY, users face a wide spectrum of tasks in their daily workflow — from binary one-tap approvals to complex, ambiguous decisions that require cross-system context and AI assistance. The problem wasn't the volume of tasks; it was that the same heavy, record-based UI pattern was being applied to all of them regardless of their actual cognitive weight.

Simple tasks like approving a job requisition or ending a work order are binary — the user already has the context they need. But navigating to a full record page to complete them added unnecessary friction. At the other end, truly ambiguous decisions like negotiating a multi-vendor SOW or handling an escalated work order dispute needed AI help to surface the right information and next action. The existing product served neither well.

A five-tier interaction system — matching UI complexity to task complexity

I designed a five-tier interaction framework that matches each UI pattern to a level of task complexity. The first three tiers keep the user entirely within the task list — no navigation, no context switching. Tiers four and five escalate the surface only when the task genuinely demands it.

#
Pattern
When to use & how it works
1
Inline Action
Row context is enough to decide
Approve / Reject CTAs directly in the row. Task resolves in place — no navigation, no modal.
Approve job before publish · End work order · Approve fixed-price payment
2
Inline Bulk
Multiple identical items pending simultaneously
Row shows item count. Clicking opens a batch tray with checklist + confirmation bar. One commit resolves all.
Multiple payment approvals · Bulk timesheet approvals · Bulk onboarding confirmations
3
Inline Expand
Row lacks context but full record isn't needed
Row expands below to reveal anomaly flags, checklists, or line items. Act without leaving the list.
Approve timesheet with anomaly · Interview coordination · Compliance checklist check
4
Modal / Canvas
Full document, multi-field flow, or multi-record review
Modal overlay or full-page canvas with structured content pane. Step-through flow with progress indicator. Save as draft always available.
PMO candidate shortlisting · Review / approve draft SOW · Work order creation · Revise job offer
5
Chat / AI Canvas
✨ Future phase — Right action is ambiguous, requires cross-system AI context
Conversational AI summarises context, surfaces cross-system data, and suggests next actions. All AI actions require explicit user confirmation before executing.
Multi-vendor SOW negotiation · Escalated work order disputes · Complex candidate shortlisting

Three phases — start simple, grow with AI

The framework is designed to ship incrementally. Each phase is complete and valuable on its own — not a stepping stone to an unfinished thing. Engineers aren't asked to build AI infrastructure upfront; they ship a clean, fast task engine first.

1
Introduce the Engine
"~75% of daily tasks completed from the homepage."
Tiers 1, 2, 3 + basic Tier 4
  • Inline actions, bulk actions, row expand
  • Snooze, re-assign, editable due dates
  • Basic modal flows for structured reviews
  • In-app notification delivery
2
Configuration & Customisation
"Tenant-level rules, ambient triggers, ecosystem automation."
Tiers 1–4 fully + rule-based proactive items
  • Stall detection, cert expiry, onboarding readiness alerts
  • Sequencing, multi-user items, auto-escalation
  • Canvas pattern for complex artifact reviews
  • Tenant configuration tools + custom item creation
3
AI & Cross-Surface
"Generative AI completes tasks; agents assist ambiguous decisions."
All tiers — Tier 5 unlocked
  • AI ambient monitoring + anomaly detection
  • AI-assisted and AI-completed items (human-in-loop)
  • Conversational Tier 5 for high-complexity tasks
  • Cross-surface delivery (Sana, Workday Everywhere, Slack)

A scalable interaction framework shipped to product & engineering

The Actions Engine framework was presented to the VNDLY product and engineering teams as the foundational system for the platform's task management experience. It gave engineers a clear, bounded scope for Phase 1 — no AI infrastructure required upfront — while establishing the full vision through Phase 3. The five-tier decision guide is now a shared reference for how the team approaches any new action type added to the platform.

5
Interaction tiers defined
28
Tasks catalogued across 6 CUJs
3
Phased rollout plan for engineering