Agent 2: Interface Archaeologist
UI/UX → User Intent Model
Role
The Interface Archaeologist examines every screen, form, report, and user interaction in the legacy system. Its goal isn't to catalog the UI — it's to reconstruct what users are trying to accomplish at each step. The intent behind the interface.
Legacy UIs are archaeological sites. Layers of additions, modifications, and workarounds accumulated over decades. The Archaeologist digs through the layers to find the original purpose.
Inputs
- Legacy application UI (screens, forms, dialogs, reports)
- User roles and permission configurations
- Navigation structures and menu hierarchies
- Output from Agent 1 (business rules catalog, to cross-reference)
Outputs
Screen Inventory
A complete catalog of every screen in the legacy system:
| Screen ID | Name | Purpose | User Roles | Business Rules Referenced |
|---|---|---|---|---|
| SCR-001 | Customer Entry | Create/edit customer records | CSR, Manager | BR-002, BR-015 |
| SCR-002 | Order Dashboard | View/manage pending orders | All | BR-003, BR-041 |
| SCR-003 | Approval Queue | Review orders requiring sign-off | Manager, VP | BR-003 |
User Journey Maps
End-to-end workflows from the user's perspective. Not "click button A, then screen B" — but "place an order, which requires customer lookup, product selection, pricing validation, and approval routing."
Each journey captures:
- The goal the user is trying to achieve
- The steps required to achieve it
- The decision points where the workflow branches
- The error states and how they're handled
- The role-based variations
Input/Output Specifications
For every form and report, a precise specification of:
- What data is collected (inputs)
- What validations are applied
- What data is displayed (outputs)
- What calculations produce the displayed values
- What triggers the display (navigation, events, schedules)
Role-Permission Matrix
A complete mapping of who can do what:
| Role | Screen Access | Data Visibility | Actions Permitted |
|---|---|---|---|
| CSR | Customer, Orders | Own region | Create, Edit |
| Manager | All | Own region | Create, Edit, Approve |
| VP | All | All regions | All |
Report Catalog
Every report the system generates — scheduled, on-demand, and ad-hoc. For each:
- What data it contains
- Who consumes it
- How often it's generated
- What business decisions it supports
Methodology
Phase 1: Surface Mapping
Catalog every screen, form, dialog, and report. Record the navigation paths between them. Build the screen inventory.
Phase 2: Intent Reconstruction
For each screen, determine the user's goal. A "Customer Entry" screen isn't about filling in fields — it's about onboarding a new customer. The intent is the requirement; the form layout is not.
Phase 3: Journey Synthesis
Connect individual screens into end-to-end user journeys. Identify the critical paths (most common workflows) and the exception paths (error handling, edge cases).
Phase 4: Cross-Reference
Map every screen interaction back to the business rules from Agent 1. Identify:
- Business rules that have no UI representation (batch/background rules)
- UI elements that enforce no business rule (cosmetic or redundant)
- Mismatches between what the UI allows and what the business rules permit
Human Validation Gate
Before passing outputs to Agent 3, the AI Software Lead:
- Confirms user journeys match actual daily workflows
- Identifies undocumented processes (workarounds users have developed)
- Validates the role-permission matrix against current organizational structure
- Flags reports that are still generated but no longer consumed