Suger / AI-Native Case Study
From ManualWorkflow Buildingto AI Orchestration
I redesigned workflow creation from node-level configuration into a guided AI-native setup flow — helping business users express intent, generate executable workflows, and refine them without starting from technical nodes.
DEMO LINK
Open prototype02 — The Problem
Business users want outcomes, not node configuration.
Business users want to complete a business process — approve a co-sell deal, send a follow-up, book a meeting — not configure a system. But traditional builders expose triggers, OAuth, field mapping, and branching logic too early. The design challenge was to reduce the technical setup burden while preserving real workflow control.
Configuration before outcome
Users must design nodes, triggers and branches before expressing what they actually want to happen.
Infrastructure exposed too early
OAuth, field mapping and API parameters surface before the business goal is even defined.
Control tied to complexity
The only way to stay in control was to understand the full technical graph — a cost most business users can't pay.
Tangled nodes · the old workflow canvas
03 — How I Reframed The Problem
I broke an ambiguous request into three product questions.
How can users describe the business workflow they want — without starting from nodes?
How can AI generate an executable workflow structure, instead of only generating text content?
How can generated workflows stay inspectable, editable, and recoverable?
04 — Design Goals
Transform workflow creation from configuring technical nodes into expressing business intent.
Reduce setup burden
Move users away from manual node configuration and toward business-level setup.
Translate intent into configuration
Use templates, setup forms, and AI parameter injection to turn business inputs into workflow logic.
Keep AI output controllable
Let users inspect, edit, activate, and refine generated workflows instead of treating AI as a black box.
05 — The Experience Model
A business setup layer before the technical builder.
This became the experience model: a business setup layer that sits before the technical workflow builder. The template defines the automation scenario. The setup form captures business intent. AI injects parameters into the right nodes. The generated workflow remains visible, editable, and activatable.
Creation pipeline
06 — Scenario-Based Setup Form
Expose business context, not infrastructure.
Decision
Expose business context instead of infrastructure settings.
Why
Users should not need to understand node parameters before defining the workflow goal.
Result
The form turns business choices into structured inputs that AI can inject into workflow nodes.
07 — Generate & Activate
Separate generation from activation.
Decision
Separate generation from activation.
Why
Users need a chance to inspect and trust AI-generated workflows before making them live.
Result
The workflow becomes executable, but still reviewable and controllable.
08 — Conversational Editing
Keep refinement inside the workflow detail page.
Decision
Keep refinement inside the workflow detail page.
Why
Users often discover changes after generation.
Result
They can adjust structure, content, and configuration through natural language — without returning to manual node setup.
Generate first. Configure and refine later.
09 — Progressive Disclosure
Structure first, details on demand.
AI generates the workflow structure first, while users complete or adjust details at the node level when needed. This keeps the creation flow fast without hiding control from advanced users.
10 — Designing For Real Product States
Designing for real product states, not just the happy path.
AI-generated workflows still need trust, recovery, and user control. So the design accounts for the states real teams actually hit — not only the ideal path from setup to activation.
Exit before activation
Leaving an unfinished workflow triggers a confirmation reminder so progress isn't lost.
Expired OAuth
Surfaces the failure clearly and offers a one-tap quick fix to reconnect.
Failed execution
Makes a broken run visible and recoverable instead of failing silently.
Setup vs. manual conflict
Reconciles conflicts between setup form updates and manual node edits.
Overwrite protection
Setup form updates warn before overwriting manual node edits.
These states mattered because AI-generated workflows still need trust, recovery, and user control.
11 — Impact & Reflection
A shift from configuration efficiency to intent expression.
Traditional workflow products optimize for configuration efficiency. This design shifts the model toward intent expression: users describe the business outcome, AI assembles the workflow, and the product keeps the result inspectable and editable.
Reflection
Users describe the business outcome. AI assembles the workflow. The product keeps the result inspectable and editable.