Internal demonstration
Fictional scenario · Harbour Apartments — Balcony Water Ingress Review
Report & Evidence Automation
A full-stack demonstration that keeps historical context, current evidence, professional findings, readiness checks, AI assistance, review and approved output inside one traceable workflow.
- Starting point
- Evidence-heavy reporting
- Workflow
- Evidence → approved report
- Authority
- Consultant + senior reviewer
End-to-end demonstration
From source material to approved professional output.
The system does more than generate text. It keeps the evidence trail, professional findings, AI assistance and review state connected as the work moves forward.
Historical context
Previous reportA previous report can be uploaded and processed. AI-extracted historical candidates keep page references and remain explicitly unverified.Current evidence
Inspection evidenceCurrent notes and photos stay inside the project with stable evidence references instead of being scattered across separate files.Human authority
Professional findingThe consultant records the current finding and connects the relevant current evidence and historical context.Deterministic check
Evidence CheckServer-owned readiness rules test whether each current finding has the required evidence structure before report work progresses.
Bounded assistance
AI Evidence AssistantEvidence questions are answered from bounded project sources with traceable references, or return an insufficient-evidence result.Factual preparation
Report draftAI can prepare a factual five-section draft from bounded project context. The consultant can review and edit the draft.Human review
Submit → review → approveThe consultant submits the work. A senior reviewer can request changes or approve; review actions do not invoke AI.Approved output
Professional preview + PDFOnly the approved version becomes the final professional preview and downloadable PDF.
Authority boundaryAI assists historical extraction, evidence questions and factual drafting. Humans own current findings, edits, submission, senior review and approval.
What this build proves
The workflow is engineered around responsibility, not just generation.
The value is not simply that AI can produce text. The useful part is keeping source material, deterministic rules and accountable people in the same operational system.
- 01Source provenance
Previous-report pages, evidence references and project context remain visible instead of becoming detached generated text.
- 02Human-authored findings
Current professional findings remain under consultant control and can be connected to the evidence that supports them.
- 03Deterministic readiness
Evidence Check and workflow rules own readiness and lifecycle decisions rather than delegating them to an AI model.
- 04Bounded applied AI
AI assists historical extraction, evidence questions and factual drafting while validation and professional authority remain visible.
- 05Review-controlled output
Submission, senior review and approval gate the final preview and PDF so generated content does not silently become an approved report.
Engineering approach
Responsibilities stay separated as the workflow moves.
The demonstration keeps business rules, evidence, bounded assistance and human approval distinct without prescribing a customer technology stack.
- 01 / InterfaceFocused operational workspace
- 02 / Business rulesServer-owned validation and workflow decisions
- 03 / EvidencePrivate, traceable project records
- 04 / Background workSeparated document-processing workflow
- 05 / AssistanceBounded AI assistance with source context
- 06 / OutputHuman-reviewed records and approved output