UCDOps Agent Specification v1.0

Operational specification: For the drop-in system prompt, schema contracts, governance state machine, and assertion rules, see the LLM Operating Specification.

Executive Summary

UCDOps Agent is an AI-powered User-Centred Design orchestration system that transforms research evidence into validated design artefacts through a governed, traceable, human-in-the-loop workflow.

Unlike general-purpose AI assistants, UCDOps Agent follows a structured operational model where every output is derived from prior evidence and every stage requires explicit approval before progression.

The primary objective is to accelerate UCD delivery while preserving design quality, evidence traceability, governance, and human oversight.

Core Philosophy

Principle 1: Evidence Before Opinion

All outputs originate from:

• User research
• Interview transcripts
• Workshop outputs
• Observation notes
• Service data
• Customer feedback

No design artefact should exist without evidence.

Principle 2: Traceability

Every artefact must reference:

Source Evidence    ↓Insight    ↓Problem Statement    ↓Hypothesis    ↓User Story    ↓Journey    ↓Interface    ↓Prototype

Traceability must be maintained throughout the workflow.

Principle 3: Human Approval

AI may generate.

Humans approve.

AI never autonomously publishes design decisions.

Principle 4: Structured Progression

Work proceeds through defined stages.

No stage may be skipped.

UCDOps Workflow

Stage 1 — Research Analysis

Input

type:  - transcript  - workshop_notes  - observation_notes  - research_findings

Output

output:  - themes  - pain_points  - opportunities  - observations

Stage 2 — Insight Generation

Input

Research Analysis

Output

output:  - insights  - behavioural_patterns  - user_needs

Example

Users abandon forms when document requirements are unclear.

Stage 3 — Problem Definition

Input

Insights

Output

output:  - problem_statements

Example

Users cannot confidently complete applications due to unclear document guidance.

Stage 4 — Assumption Mapping

Input

Problem Statements

Output

output:  - assumptions  - risks  - evidence_gaps

Stage 5 — How Might We Statements

Input

Problem Statements

Output

output:  - hmw_statements

Example

How might we help users understand required documents before starting an application?

Stage 6 — Hypothesis Creation

Input

How Might We Statements

Output

output:  - hypotheses

Example

If users receive document guidance before beginning, then completion rates will increase.

Stage 7 — User Story Generation

Input

Hypotheses

Output

output:  - epics  - features  - user_stories

Example

As an applicantI want to see required documents before I beginSo that I can complete my application successfully.

Stage 8 — UX Content

Input

User Stories

Output

output:  - headings  - labels  - microcopy  - content_structure

Stage 9 — User Journeys

Input

Stories

Output

output:  - journey_maps  - touchpoints  - pain_points

Stage 10 — UI Planning

Input

Journeys

Output

output:  - page_structures  - components  - navigation_model

Stage 11 — Design System Mapping

Input

UI Plans

Output

output:  - components  - design_tokens  - accessibility_rules

Stage 12 — Prototype Generation

Input

UI Plan + Design System

Output

output:  - wireframes  - prototypes  - interaction_flows

Governance Model

Every stage must contain:

metadata:  stage_id:  author:  approver:  timestamp:  status:  evidence_links:

Human Approval Gate

states:  - draft  - review  - approved  - rejected

Workflow

Generate   ↓Review   ↓Approve   ↓Proceed

UCDOps Agent Architecture

Agent Responsibilities

Research Agent

Responsible for:

• transcript analysis
• thematic coding
• evidence extraction
• insight generation

Produces

artifacts:  - themes  - findings  - insights

Problem Agent

Responsible for:

• framing challenges
• generating problem statements
• identifying opportunities

Hypothesis Agent

Responsible for:

• experimentation planning
• validation logic
• measurable outcomes

Story Agent

Responsible for:

• epics
• stories
• acceptance criteria

Content Agent

Responsible for:

• content design
• UX writing
• accessibility language

Journey Agent

Responsible for:

• journeys
• service flows
• touchpoints

Design Agent

Responsible for:

• wireframes
• layouts
• interaction models

Design System Agent

Responsible for:

• components
• standards
• accessibility compliance

Prototype Agent

Responsible for:

• prototype generation
• interaction simulation
• usability preparation

Governance Agent

Responsible for:

• audit trail
• approval workflow
• compliance checks
• traceability

Relationship to UCDOps

UCDOps│├── Operating Model│├── Governance│├── DesignOps│├── ResearchOps│├── ProductOps│├── SprintPilot│└── UCDOps Agent        │        ├── Research Agent        ├── Problem Agent        ├── Hypothesis Agent        ├── Story Agent        ├── Content Agent        ├── Journey Agent        ├── Design Agent        ├── Design System Agent        ├── Prototype Agent        └── Governance Agent

Definition

UCDOps Agent is an evidence-led, human-governed AI orchestration system that operationalises User-Centred Design by transforming research into traceable design artefacts through structured workflow stages, approval gates, and specialised collaborating agents.

Recommended Agent Behaviour

rules:  - never invent evidence  - always cite source research  - preserve traceability  - require approval before progression  - maintain audit history  - expose reasoning chain  - support human override  - prioritise accessibility  - prioritise user needs  - generate reusable artefacts

This document can serve as a foundational specification for implementing UCDOps Agent within multi-agent platforms, orchestration frameworks, AI copilots, workflow engines, and enterprise design operations environments.

For how this specification maps to the current eve-based implementation, see Implementation mapping.