Application

From using AI tools, observations and research, I identified areas I wanted to explore through my design practice.

Rather than building a production tool, I focused on testing how these principles might translate to real design decisions through a conceptual stakeholder synthesis tool called Scaffold.

Context

Scaffold shows how AI can accelerate stakeholder synthesis from weeks to minutes without removing human judgment through transparent, validated collaboration.

This conceptual tool let me practice applying transparency and validation principles to a real design problem I've experienced.

The Problem

AI that isn't explicit with its reasoning which can make user trust challenging. They aren't able to validate outputs or understand where AI might be wrong.

The Principle

Be clear with AI reasoning, confidence levels, and evidence at every decision point so users can verify claims and maintain appropriate doubt.

Implementation in Scaffold

• Confidence scores
• Evidence including number of mentions, quotes, and direct from source materials
• Validation required - user reviews before AI proceeds

The Problem

AI is designed to replace human work which removes the thinking that makes the work valuable. This leads to an over-reliance on potential inaccurate recommendations.

The Principle

AI handles pattern recognition at scale whilst users make strategic judgment calls.

Implementation in Scaffold

• AI generates strategic options with explicit varying degrees of confidence
• Each option shows stakeholder fit percentage, pros/cons, trade-offs
• AI recommends whilst user compares and decides
• AI never auto-selects or advances without the user decisions

Transparency vs. Simplicity

Progressive disclosure of information allows for a balances between summaries for scanning and expandable details for validation

Speed vs. Agency

Chose not to auto select the option with highest confidence, the user always chooses the strategic direction to maintain control

Confidence Communication

Maintained transparency and honesty by keeping low confidence scores

This conceptual project let me practice applying AI/UX principles through essential workflows, not comprehensive features.

A production version would expand functionality, but designed around the same core principles.

Transparency needs additional supporting behaviours

Through this project, I learned that transparency needs validation loops and confidence communication to be actionable - these principles work as a feedback system.

Restraint over feature overload

This exploration taught me that effective AI/UX comes from designing the minimum interactions that maintain human control, not exposing every capability the AI can perform.