Data Flow Sequence | Project Northern Lights

🗣 Copilot Conversational 👤 Human Approval Required

Analyst Initiates Query

User asks question or requests action via Copilot

👤 Human

Copilot Chat Interface

👤

"What's the status of the Environmental section? Are all required energy inputs complete?"

🗣

Query Parsing

Copilot interprets intent and identifies required tools

🔧

Tool Selection

Routes to check_completion tool endpoint

↓

Document Analysis

MCP Tool Server analyzes uploaded documents

🤖 AI

📄

analyze_document

Extract KVPs from Energy_Data_2025.xlsx using existing extraction logic

✅

check_completion

Validate against VSME Environmental section requirements

Extracted Data (KVPs)

{
  "section": "VSME Environmental",
  "entity_name": "KPMG Norway AS",
  "energy_data": {
    "total_consumption_kwh": 1245000,
    "renewable_percentage": 78.5,
    "scope_1_emissions": 12500,
    "scope_2_emissions": 8900
  },
  "completion_status": {
    "required_fields": 12,
    "completed_fields": 10,
    "missing": ["water_usage", "waste_recycled"]
  }
}

↓

Status Response

Copilot presents findings to analyst

🤖 AI

Copilot Response

🤖

Environmental Section Status: 83% Complete

✅ Energy consumption data: Complete
✅ Renewable percentage: 78.5%
✅ Scope 1 & 2 emissions: Complete
❌ Missing: Water usage, Waste recycled

Would you like me to draft the narrative for the completed energy data?

↓

Narrative Draft Request

Analyst requests draft generation

👤 Human

Analyst Request

👤

"Yes, please draft the narrative for energy consumption. Use a professional tone matching our previous reports."

📝

draft_narrative

Invoke narrative generation with extracted KVPs

🤖

Azure AI Foundry

Generate VSME-compliant narrative using reasoning model

Generated Draft

{
  "narrative_text": "In fiscal year 2024, KPMG Norway AS demonstrated
continued commitment to sustainable energy practices. Total
energy consumption reached 1,245 MWh, with renewable sources
accounting for 78.5% of the total energy mix—an improvement
of 5.1 percentage points from the prior year.

Direct emissions (Scope 1) totaled 12,500 tCO2e, while
indirect emissions from purchased energy (Scope 2) were
8,900 tCO2e. The organization continues to prioritize
energy efficiency initiatives and renewable energy
procurement.",
  "confidence_score": 0.94,
  "vsme_compliance": "Aligned with E1-5 disclosure requirements"
}

↓

Analyst Review & Approval

Human decision gate before write-back

👤 Human

Draft Presented for Review

🤖

Draft Narrative (Energy Consumption)

"In fiscal year 2024, KPMG Norway AS demonstrated continued commitment to sustainable energy practices..."

Confidence: 94% | VSME Aligned: E1-5

👤 Human Approval Required

Analyst must explicitly approve before writing to Workiva

✅ Approve & Write 📝 Request Revision

Analyst Approval

👤

"The draft looks good. Please write it to Workiva."

↓

Controlled Write-back

Approved content written to Workiva

🤖 AI

🔐

Key Vault

Retrieve Workiva OAuth credentials securely

🔗

write_workiva

Execute PUT request to VSME Environmental section

📄

Workiva API

Update target container with approved narrative

Workiva API Request

PUT /platform/v1/spreadsheets/{id}/sheets/{id}/data
Authorization: Bearer {access_token}

{
  "cells": [
    {
      "row": 15,
      "column": "B",
      "value": "In fiscal year 2024, KPMG Norway AS..."
    }
  ]
}

Confirmation

🤖

✅ Successfully written to Workiva

Section: VSME Environmental (Energy Consumption)
Status: Draft Ready for Review
Timestamp: 2025-01-15 09:45:32 UTC

Open in Workiva →

📊 Full Audit Trail

Every step is logged to Azure Log Analytics: tool invocations, AI prompts/responses, analyst decisions, and Workiva write operations. This ensures complete auditability and supports GDPR/EU AI Act compliance requirements.

👤

Human Action

🔧

Tool Execution

🤖

AI Processing

👤

Approval Gate

📄

External API