woodysigner.
PT
← Back to projects

Case 04 / Complementary case / AI / Review criteria

AI-first & Prompt Thinking
Method and experimentation guided by design judgment

A view into my experimentation process: frame a problem, formulate instructions, explore alternatives and review the result. Artificial intelligence (AI) supports execution; design criteria guide choices.

Contribution

Prototype structuring and review

Scope

Instructions · Prototypes · Review

Status

Documented experimental process

Evidence in the material

Context and challenge

Prototype histories record instructions to preserve components, correct navigation and refine hierarchy. They provide specific examples of bounded changes and acceptance criteria.

Role and collaboration

My role is to formulate context, guide alternatives and review output. Domain validation and technical feasibility depend on specialists and engineering. Instructions are not evidence of completed research or validation.

Relevant constraints

An instruction may be only partially executed and plausible output may contain errors. Intent, prototype behavior and implementation must be distinguished. This process does not replace user research or technical review.

Guiding problem

How can AI support exploration without taking over design judgment?

Design decisions

Proposed design decision

Bound the change

Income Tax instructions request reusing Tax Brain spacing, components and patterns, connecting the experiment to an established experience.

Proposed design decision

Specify interactions and acceptance criteria

Tax Brain instructions require list and timeline task details to use the same fields. Criteria cover correct selection, consistency and updates in both views.

Proposed design decision

Review hierarchy instead of adding noise

An Income Tax iteration asks for fiscal-impact row emphasis and neutral financial values, prioritizing impact, alert and net value while retaining the table structure.

Proposed design decision

Declare demonstration limits

Trade Flow chat uses predefined responses. Describing it as a simulation keeps claims proportional to the evidence.

Proposed design decision

Journey and flow

  1. Define problem and usage context
  2. Bound scope and patterns to preserve
  3. Write interactions and acceptance criteria
  4. Generate and explore a prototype
  5. Review behavior, content and hierarchy
  6. Document limits and plan validation

Screens and experience

Documented example · Income Tax

Refine fiscal-impact visibility without changing the table structure.
Preserve column order, filters and comment behavior.
Highlight rows with impact; keep financial values neutral.
Check signs, typographic hierarchy and readability.

Summary of instructions found in the history; this is not a validation result.
Evidence in the material

Documented iterations

Income Tax: consistency

The first accessible instruction requests continuity with Tax Brain; a later one specifies a shared global header.

Income Tax: reading hierarchy

The history records a table refinement prioritizing fiscal impact and using signs and typography for financial values.

Tax Brain: navigation criteria

The side-panel instruction lists opening, field consistency, task creation and editing checks.

Validation plan

Status and validation

Documented experimental process

The examples demonstrate instructions and intended review. No measured productivity gains were supplied. The proposed evaluation records review time, criterion deviations, rework and decision validation with users and engineering.

Hypothesis to validate

How can AI support exploration without taking over design judgment?

Learnings

01 — Specific instructions make deviations easier to identify.

02 — Preserving an existing system may matter more than generating a new screen.

03 — Acceptance criteria describe what to review; they do not prove the solution passed that review.

Case sources

  • Income Tax (Automation Anywhere).make — consistency and table-refinement instructions
  • Tax Brain PROCUREMENT.make — interaction instructions and acceptance criteria
  • v9_10_daycoval_trade_flow_demo_safe.html — simulated assistance logic