Project Aura · Final Deliverable Screens

The Product.
What Analysts Actually Use.

Beyond the process and methodology, this is what was built. Six production-grade UI screens showing the complete analyst journey — from IA through final compliance export. Every component, data state, and human-AI handoff moment is documented here.

Information Architecture

→

User Flow

→

Dashboard

→

AI Analysis

→

ESG Report

→

Human Approval

→

Audit Trail

Section A · Information Architecture

Aura IA — Navigational Hierarchy

The IA is structured around three primary user roles: Analyst (extraction & review), Lead (oversight & pipeline), Compliance (audit & export). All paths converge at the Human Approval Gate before any data touches the legacy database.

◈ AURA COPILOT

Root Application Shell

📊 Dashboard

Analyst Home · Live Queue

🔍 Analysis

Document Processing

✅ Review

Human Approval Gate

📄 Reports

AI-Generated ESG Output

🔒 Audit Trail

GDPR Compliance Log

⚙️ Settings

Admin · Integrations

Audit Queue

Live · Priority Sort

Team Pipeline

Thomas Lead View

AI Performance

Confidence Metrics

Upload PDF

Drag · Drop · Bulk

AI Extraction

Copilot Side Panel

Source Viewer

Inline Annotation

Approve / Flag

Field-Level Control

Manual Override

Edit + Re-cite

Push to Legacy

API Write Gate

ESG Summary

Auto-Generated

CSRD Mapping

Regulation Check

Export PDF

Regulator Ready

Decision Log

Immutable · Timestamped

GDPR Export

Julian Compliance View

Data Lineage

Source → Decision

RAG Config

Threshold Control

Legacy DB API

Write Permissions

User Roles

Analyst · Lead · Compliance

IA Design Decision

Review is not optional — it is architecturally required

The IA physically prevents a direct path from AI Extraction to Legacy Push. Every analyst must pass through the Review node — this is not a UX nudge, it's a structural constraint that satisfies Julian's GDPR requirement and eliminates the possibility of automated data write without human sign-off.

Roles Mapped

3

Core Screens

6

Mandatory Gates

1

Section B · User Flow

Analyst Journey — Upload to Commit

The complete end-to-end flow from when an analyst receives an audit assignment to when the structured ESG data is written to the legacy insurance database. Every AI decision is surfaced, every human intervention is explicit.

Phase 1 — Intake & Upload

Step 01

Receive Audit Assignment

Sarah receives client utility bill audit via Aura dashboard notification. Client: AUR-EU-4471. Priority: High. Deadline: 48hrs.

Sarah · Analyst

Open

Step 02

Upload Client PDF

Drag-drops 50-page utility bill PDF into Aura. Document is processed: indexed, chunked, embedded into the internal RAG vector store.

Sarah · Analyst

Auto-process

Step 03

RAG Ingestion

AI Gateway indexes document chunks. Emission tables, consumption data, and CSRD thresholds are mapped to semantic embeddings. No external API calls.

AI System

Ready

Step 04

Document Ready Signal

Aura surface signals "Document indexed — 94 extractable fields identified." Analysis workspace activates with document viewer and Copilot panel.

System

Phase 2 — AI Extraction

Step 05

Query Copilot

Sarah types: "Extract all Scope 1, 2, and 3 emissions data and compare against our sector benchmark database." NLP router classifies as Extraction + Comparison intent.

Sarah · Analyst

NLP classify

Step 06

Multi-Agent Retrieval

Extraction Agent pulls data from uploaded PDF. Comparison Agent retrieves sector benchmarks from internal legacy DB. Both agents return structured data cards.

AI Agents × 2

Surface results

Step 07

Explainability Output

Each extracted value shows: value, confidence score, source document page, paragraph reference, and benchmark delta. Confidence below 85% triggers yellow flag automatically.

AI Output

Present

Step 08

Review Extracted Fields

Sarah reads each field, spot-checks 3 citations by clicking the source link (highlights exact PDF paragraph). Identifies one value needs correction — uses manual override.

Sarah · Analyst

Phase 3 — Human Approval Gate

Step 09

Manual Override

Sarah edits Scope 2 value from AI-suggested 893 to correct 847. Provides re-citation: selects correct paragraph from PDF viewer. Override logged with analyst ID and timestamp.

Sarah · Override

Field decision

Decision Gate

Approve All Fields?

Field-level: Approve ✓ Flag ✗ Override ✎

All approved

Step 10

Approve & Generate Report

Sarah clicks "Approve All & Generate ESG Report." Aura compiles all approved fields into structured ESG summary with CSRD compliance check run automatically.

Sarah · Approved

Auto-compile

Step 11

Push to Legacy DB

Structured data card is schema-validated and written to the legacy CIS database via API. Audit log entry created. Thomas and Julian receive compliance confirmation notification.

API Write · Logged

🔐

Zero Data Leaves the Platform

Every RAG query in Steps 06 and 07 is resolved entirely within the Azure private cloud. The PDF never leaves the client's infrastructure.

✋

Human Must Touch Every Field

Step 09 ensures even AI values Sarah "agrees with" require an explicit Approve action — preventing passive rubber-stamping and maintaining the override audit log.

📋

Every Action is Append-Only Logged

Steps 09 and 11 generate immutable audit events. Julian can reconstruct the complete AI-to-decision chain for any record at any time for regulator submission.

Screen 01 · Analyst Dashboard

Live Audit Pipeline
Thomas's Oversight View

The dashboard gives every stakeholder their primary signal at a glance. Sarah sees her queue. Thomas sees team throughput. The AI performance metrics are always visible — not buried in settings.

Screen 01 / 05 · Dashboard · Role: All Users Primary landing screen after login. Adapts content columns based on role permissions. Sarah sees personal queue; Thomas sees team-wide pipeline.

aura.cis-platform.internal / dashboard

🔒 Secure Instance

12

Audits In Queue

↑ 3 urgent

1.4h

Avg Processing Time

↓ 65% vs baseline

94%

AI Confidence Score

Last 7 days avg

14%

Human Override Rate

Within target band

Active Audit Queue View All →

AUR-EU-4471 · Utility Bill Q3

ProTech Industries

Urgent

2h left

AUR-EU-4489 · ESG Report FY24

NordikStahl GmbH

In Review

8h left

AUR-EU-4502 · CSRD Compliance

Solaris BV Amsterdam

AI Processing

24h left

AUR-EU-4477 · Scope 3 Analysis

BavarianMotors AG

Pending Sign-off

48h left

AUR-EU-4481 · Premium Calc

Renova SpA

Approved

Done ✓

AI Performance · Week

Extraction Accuracy by Type

Scope 1

Scope 2

Scope 3

Energy

Water

Recent Activity

Sarah M. approved 14 fields · AUR-EU-4477

4m

AI flagged anomaly · Scope 2 delta +22% · AUR-EU-4489

17m

New PDF indexed · 94 fields identified · AUR-EU-4471

31m

Report generated · AUR-EU-4461 · Pushed to Legacy DB

1h

Screen 02 · AI Analysis Workspace

Document Analysis +
AI Copilot Side Panel

The core working screen. Sarah sees the actual client PDF on the left, highlighted by AI extraction annotations. The Copilot panel on the right shows every extracted value with its source citation, confidence, and inline approve/flag controls. This is where trust is built or broken.

Screen 02 / 05 · Analysis Workspace · Role: Analyst Split-view: live PDF document viewer (left) with AI Copilot extraction panel (right). Source citations are clickable and scroll the PDF to exact paragraph.

aura.cis-platform.internal / analysis / AUR-EU-4471

🔒 Secure Instance

ProTech Industries AG

Annual Environmental Data Report · FY 2024 · Q3 Supplement

Page 7 / 50
Ref: PTI-ENV-2024-Q3

Section 3.2 — Greenhouse Gas Emissions Inventory

Reporting Period:01 Jan 2024 – 30 Sep 2024

Methodology:GHG Protocol Corporate Standard (2023)

Scope 1 — Direct Emissions

Natural gas combustion:189.4 tCO₂e

Company vehicle fleet:122.6 tCO₂e

Total Scope 1:312.0 tCO₂e

Scope 2 — Indirect Energy Emissions

Electricity purchased (market-based):847.2 tCO₂e

District heating:64.3 tCO₂e

Renewable energy offset:–88.0 tCO₂e

Net Scope 2 (market-based):823.5 tCO₂e

Annual Energy Consumption

Energy Type	Consumption	Unit	YoY Δ
Grid electricity	8,420	MWh	+2.1%
Natural gas	3,180	MWh	-3.4%
Renewable (self-gen)	840	MWh	+34.2%
Total	12,440	MWh	+0.8%

⚠ Note: Scope 2 market-based figure reflects contractual instrument adjustments per CSRD Article 29b requirements. See Appendix C for EAC certificate details.

Aura Copilot

RAG · Internal Sources Only

Sarah · Query

Extract all emissions data and compare against our sector benchmark for manufacturing.

Aura · Extraction Complete

Found 7 emission values across Scope 1, 2, and energy data. Scope 2 market-based is 14.5% above manufacturing sector benchmark.

Scope 1 Total 312.0 tCO₂e

Scope 2 Market 847.2 tCO₂e

Benchmark Delta +14.5% ▲

Total Energy 12,440 MWh

Renewable Ratio 6.75%

Source: 📎 Bill_Q3_2024.pdf · p.7, §3.2

Benchmark: 📎 Legacy DB · Sector_Mfg_2024

Confidence

94%

Aura · CSRD Alert

⚠ Scope 2 exceeds CSRD Article 29b disclosure threshold for Category B manufacturers. Manual CSRD annotation required before export.

📎 CSRD Regulation (EU) 2022/2464 · Art.29b

✓ Approve All

✎ Override

✕ Flag

Screen 03 · AI-Generated ESG Report

Final ESG Report Output
Structured · Cited · Regulation-Mapped

Once Sarah approves all fields, Aura auto-compiles the structured ESG report. Every finding is pinned to its source document. CSRD compliance status is mapped field by field. Julian can export the entire report as a regulator-ready PDF with full decision trail embedded.

Screen 03 / 05 · ESG Report View · Role: Analyst + Compliance AI-compiled summary of all approved extracted data. Right sidebar shows real-time CSRD compliance mapping. Every metric is source-cited and human-verified before this screen generates.

aura.cis-platform.internal / reports / AUR-EU-4471 / esg-summary

🔒 Secure Instance

ProTech Industries AG · ESG Audit Report

Ref: AUR-EU-4471

Period: Q3 FY2024

Analyst: Sarah M.

Status: ✓ Human Approved

Generated: 14 Nov 2025 · 14:32

Source Fields: 94 extracted · 94 approved · 1 overridden

847.2

Scope 2 Emissions
tCO₂e · Market-Based

312.0

Scope 1 Emissions
tCO₂e · Direct

12,440

Energy Consumption
MWh · Total Annual

6.75%

Renewable Energy
Ratio · Self-Generated

Key Findings — AI Extracted & Human Verified

🔴

Scope 2 Emissions 14.5% Above Sector Benchmark

Market-based Scope 2 emissions of 847.2 tCO₂e exceed the manufacturing sector benchmark of 740 tCO₂e by 107.2 tCO₂e (+14.5%). This triggers a CSRD Article 29b disclosure requirement and will affect the premium calculation band. Client should be engaged on renewable procurement strategy for FY2025.

📎 Bill_Q3_2024.pdf · p.7, §3.2 · Legacy DB · Sector_Mfg_2024

🟡

Renewable Energy Ratio Below SFDR Threshold

Self-generated renewable energy at 6.75% is below the SFDR PAI indicator threshold of 15% for Article 8 fund eligibility. This client is currently classified as Category B. Increasing to 840+ MWh self-generation would unlock Category A premium band (est. saving: €12,400/yr).

📎 Bill_Q3_2024.pdf · p.7, Table 3.1 · SFDR Reg (EU) 2019/2088

🟢

Scope 1 Emissions Declining — Positive Trend

Direct Scope 1 emissions reduced by 3.4% YoY (from 323.2 tCO₂e in FY2023 to 312.0 tCO₂e in Q3 2024 annualised). Fleet electrification programme noted in Appendix A is the primary driver. This should be reflected positively in the premium narrative.

📎 Bill_Q3_2024.pdf · p.7, §3.2 · PrevAudit_2023.pdf · p.4

AI Report Compiled
All fields human-verified

CSRD Compliance Map

Art.29b Scope 2 Threshold

Exceeded

Art.29b Renewable Ratio

At Risk

Scope 1 GHG Protocol

Compliant

SFDR PAI Disclosure

Filed

EU Taxonomy Alignment

Pending

Premium Impact Estimate

Cat B

Risk Band · Current

Scope 2 breach triggers +2.3% premium loading per actuarial table ESG-2024-B. Recommend client advisory on renewable procurement.

Julian's Export

Full GDPR-compliant decision trail ready. 94 events logged.

⬇ Export Audit PDF

Screen 04 · Human Control & Approval Interface

Field-Level Human Control
The Approval Gate — Sarah's Most Critical Screen

This is where trust is operationalized. Sarah reviews each AI-extracted field individually. She can Approve (accept AI value), Override (edit with re-citation), or Flag (reject and escalate). No field reaches the legacy database without passing through this screen with an explicit human decision attached.

Screen 04 / 05 · Review & Approval · Role: Analyst (Required) The human-in-the-loop gate. Left panel: AI-extracted data fields with source links and confidence indicators. Right panel: original source document context. Bottom bar: batch approval with full audit log preview.

aura.cis-platform.internal / review / AUR-EU-4471 / approval

🔒 Secure Instance

Scope 2 & Energy Fields — AI Extraction Review

Client: ProTech Industries AG

Source: Bill_Q3_2024.pdf

AI Confidence: 94%

AI Extracted Values — Awaiting Decision 8 fields

Scope 1 Total

312.0 tCO₂e ✓ Approved

📎 p.7

Scope 2 Market-Based

Overridden

893.0 tCO₂e (AI)

847.2 tCO₂e (Sarah)

📎 p.7

Net Scope 2 (Market)

823.5 tCO₂e

📎 p.7

✓ / ✕

Total Energy Consumption

12,440 MWh

📎 p.7

✓ / ✕

Renewable Self-Gen

840 MWh

📎 p.7

✓ / ✕

Renewable Energy Ratio

6.75% ⚠ Below SFDR 15% threshold

📎 p.7

Benchmark vs Sector

+14.5% above Mfg avg

📎 Legacy DB

✓ / ✕

CSRD Article 29b Flag

⚠ Threshold exceeded · requires annotation

📎 CSRD 2022

✓ / ✕

Source Context — Active Citation 📎 p.7 §3.2

Override Justification — Sarah M.

AI extracted 893.0 tCO₂e from Table 3.1 Row 2. However, correct value is 847.2 from §3.2 narrative text which reflects contractual instrument adjustments per CSRD Art.29b. Table row is location-based, not market-based. Paragraph value supersedes table row for CSRD compliance purposes.

Corrected citation: Bill_Q3_2024.pdf · p.7, §3.2 · Para 2 · Line 3

§3.2 — Scope 2 Emissions (Market-Based Method)

For the reporting period January – September 2024, ProTech Industries AG reports Scope 2 greenhouse gas emissions calculated using the market-based method in accordance with the GHG Protocol Corporate Accounting and Reporting Standard.

Electricity purchased (market-based): 847.2 tCO₂e — reflecting contractual instruments (EACs) as per CSRD Article 29b §4(c) requirements. This figure supersedes location-based calculations where EAC documentation is available. See Appendix C for certificate registry numbers.

Note: Location-based figure of 893.0 tCO₂e shown in Table 3.1 is retained for reference only and must not be used for CSRD Article 29b disclosure.

Screen 05 · Compliance Audit Trail

Immutable Decision Log
Julian's GDPR Compliance View

Every AI query, every extraction, every human approval, every override, every push to the legacy database — logged, timestamped, attributed. Julian can filter, search, and export a complete decision trail for any audit within seconds. This is the legal backbone of the entire system.

Screen 05 / 05 · Audit Trail · Role: Compliance Officer (Julian) Append-only log of every AI and human action across the platform. Fully exportable as a GDPR-compliant PDF. Cannot be edited, deleted, or backdated by any user role including admin.

aura.cis-platform.internal / audit-trail / AUR-EU-4471

🔒 Secure Instance · GDPR Compliant

All Events

AI Actions

Human Decisions

Overrides

System Events

⬇ Export · GDPR Art.5 Compliant PDF

🔐 Append-only log · 94 events · No modification possible · Data controller: CIS Division · Retention: 7 years per GDPR Art.5(1)(e)

Timestamp	Event Type	Actor	Field / Action	Value	Source Citation	Confidence
14 Nov · 14:32:07	DB WRITE	Sarah M.	All 94 fields → Legacy DB	Committed ✓	API v2.4 · Schema validated	—
14 Nov · 14:31:54	APPROVED	Sarah M.	All remaining fields batch approval	61 fields	Bill_Q3_2024.pdf · multi-page	94%
14 Nov · 14:28:41	OVERRIDE	Sarah M.	Scope 2 Market-Based Emissions	893.0 tCO₂e (AI) 847.2 tCO₂e (human)	Bill_Q3_2024.pdf · p.7 §3.2 ¶2	N/A Override
14 Nov · 14:24:19	AI EXTRACT	Aura AI	Scope 2 Market-Based Emissions	893.0 tCO₂e	Bill_Q3_2024.pdf · p.7 Tbl 3.1	71%
14 Nov · 14:24:18	AI EXTRACT	Aura AI	Scope 1 Direct Emissions	312.0 tCO₂e	Bill_Q3_2024.pdf · p.7 §3.2 ¶1	97%
14 Nov · 14:23:09	AI QUERY	Sarah M.	Natural language query submitted	"Extract all emissions data…"	Internal RAG · No external calls	—
14 Nov · 14:18:43	PDF INDEXED	System	Document ingested into RAG vector store	94 fields identified	Azure Private Cloud · EU-West	—
14 Nov · 14:18:01	UPLOAD	Sarah M.	PDF uploaded to secure workspace	Bill_Q3_2024.pdf · 50pp · 3.2MB	No external transfer · GDPR Art.5	—

⚖️

Every Row is Append-Only

The audit log uses an append-only data model. Not even admin users can modify or delete rows. This is enforced at the database level, not just the UI — satisfying Julian's requirement for a truly immutable record.

🔬

Confidence Score Preserved

The AI's original confidence score (71% for the overridden Scope 2 value) is preserved alongside the human correction. This is evidence of the Explainability Seam working — Sarah questioned a low-confidence extraction and was right to do so.

📤

One-Click Regulator Export

Julian can export all 94 events as a formatted PDF within seconds. The export includes the source document references, the AI confidence scores, and each human decision with its justification text — ready for EU regulator submission.

The Product.What Analysts Actually Use.

Aura IA — Navigational Hierarchy

Analyst Journey — Upload to Commit

Live Audit PipelineThomas's Oversight View

Document Analysis +AI Copilot Side Panel

Final ESG Report OutputStructured · Cited · Regulation-Mapped

Field-Level Human ControlThe Approval Gate — Sarah's Most Critical Screen

Immutable Decision LogJulian's GDPR Compliance View

The Product.
What Analysts Actually Use.

Live Audit Pipeline
Thomas's Oversight View

Document Analysis +
AI Copilot Side Panel

Final ESG Report Output
Structured · Cited · Regulation-Mapped

Field-Level Human Control
The Approval Gate — Sarah's Most Critical Screen

Immutable Decision Log
Julian's GDPR Compliance View