EU AI Act Compliance and Risk Management ReportBook for ML Deployments

A comprehensive PowerBI solution that transforms complex AI compliance data into actionable insights for ML deployments. The dashboard integrates multiple data sources tracking model performance, compliance scores, bias detection, regulatory deadlines, and user appeals. Features include real-time compliance monitoring, automated risk assessment, bias tracking, and performance metrics visualization. Essential for organizations deploying ML models under EU AI Act regulations, providing a unified view for compliance, risk management, and regulatory reporting.

SQL Server Power BI

The following guide provides step-by-step instructions for creating a studies management software in Excel with the input data visualized in Power BI. The presented data is exported from Excel and the visualizations in PowerBI will update as changes to cell data are made in Excel.

Some key features include:

◍ The ability to manage studies in a very granular and in-depth manner
◍ The ability to modify the platform to accommodate a wide variety of needs and situations
◍ Powerful visualization and insight capabilities
◍ Interactive and easily understandable user interface

Below is an example of the finished studies manager when loaded into PowerBI.

Try the different functions by interacting with the elements on the canvas. By pressing this icon you can go into full-screen mode. Switch between the different pages of the report here

Press ESC to return to this page.

Guide for Creating Project

PHASE[1]
Project Setup & Requirements

Before diving into the technical implementation of the AI compliance monitoring system, it’s crucial to establish a solid foundation for the project. This phase focuses on defining the fundamental elements that will ensure our PowerBI ReportBook effectively tracks and manages AI compliance under the EU AI Act framework.

During this phase, we’ll outline specific compliance monitoring objectives, set up the development environment with the necessary tools, plan the data architecture, and identify data sources that will feed into the compliance tracking system. This sets the foundation for building a robust solution that can effectively monitor model performance, compliance scores, bias detection, regulatory deadlines, and other features.

Define the overall objectives

In this project, our primary objective is to build a robust AI compliance monitoring system that aligns with the EU AI Act. Clearly defining our goals will ensure that the Power BI ReportBook matches our expectations.

Key Objectives for This Project

Ensure Compliance with the EU AI Act
We need to track regulatory deadlines, risk levels, and legal obligations, ensuring continuous monitoring and reporting for AI model compliance.
Implement Risk & Bias Monitoring
Our system should identify and flag potential biases in AI decisions, assess risk levels, and provide automated alerts for compliance deviations.
Enable Performance & Auditability
The ReportBook should provide real-time compliance scores, audit logs, and historical tracking, making it easy to review past compliance data.
Support Stakeholder Needs
We must design custom dashboards, alerts, and access controls to meet the needs of compliance officers, risk managers, and data teams.

By setting these objectives, we establish a clear direction for building a scalable and efficient AI compliance monitoring system in Power BI.

Install required tools

Since our project requires a controlled environment for data processing and compliance monitoring, we will use VMware to set up a Windows Server virtual machine (VM). This VM will host both SQL Server and Power BI Desktop, ensuring seamless integration and live query capabilities.

To make the project as easily accessible as possible, we will use only free versions of the different required software.

VMware Setup

Download VMware Workstation Player (free version):

https://www.vmware.com/products/workstation-player.html
Install VMware:
- Run the downloaded .exe file
- Follow the installation wizard
- Select “Free for non-commercial use” option

Windows Server Setup

Download Windows Server 2022 Evaluation (free 180-day trial):

https://www.microsoft.com/en-us/evalcenter/evaluate-windows-server-2022
Create new VM in VMware:
- Click “Create a New Virtual Machine”
- Select “Custom (Advanced)”
- Mount the Windows Server ISO
- Allocate resources:
  - 8-16GB RAM
  - 114-256GB storage
  - 2-4+ CPU cores
- The more resources you allocate the more efficient the VM will be
  - The above are resource allocation suggestions
  - Allocated resources are released if the VM is not in use
Install Windows Server:
- Choose “Windows Server 2022 Standard (Desktop Experience)”
- Follow installation wizard
- Install VMware Tools after OS installation

SQL Server Setup (On VM)

Download SQL Server Developer Edition (free for development):

https://www.microsoft.com/en-us/sql-server/sql-server-downloads
Install SQL Server:
- Run installation wizard
- Select “Custom” installation
- Install Database Engine Services and Management Tools
Configure SQL Server:
- Enable mixed mode authentication
- Set up SA password
- Enable TCP/IP protocol
- Configure firewall rules if needed

SQL Server Management Studio (SSMS) Setup

Navigate to the official Microsoft download page for SSMS.
Download the latest SQL Server Management Studio installer (it’s free).
Run the installer on your Windows Server VM and follow the default prompts.
Once installed, launch SSMS and connect to your SQL Server instance using the credentials you configured earlier.

This ensures you have a fully featured interface for interacting with your SQL Server database – essential for the importing, querying, and administration steps in the next sections.

Power BI Desktop Setup (On VM)

Download Power BI Desktop (free version):

https://powerbi.microsoft.com/desktop/
Install on Windows Server VM:
- Run installer
- Follow setup wizard
- Sign in using an organizational (work/school) account (Power BI requires authentication for cloud services).
Configure for local SQL Server:
To connect Power BI to our SQL Server database on the VM
- Launch Power BI Desktop
- Click Home → Get Data → SQL Server
- Enter the SQL Server name (e.g., localhost or VM-SQLServer)
- Select the appropriate data connectivity mode
  - Import Mode: Loads data into Power BI
  - DirectQuery Mode: Keeps data in SQL Server
- Authenticate using SQL credentials (use the SA account if configured)
- Select the database (e.g., Compliance_DB) and load the data

Plan the architecture

System Overview

The AI compliance monitoring system will be deployed on a Windows Server virtual machine (VM), hosting both SQL Server and Power BI Desktop. The architecture consists of the following components:

Data Sources: Structured and unstructured compliance-related data (e.g., model performance metrics, risk scores, user appeals, and regulatory deadlines) collected in CSV format.
Database Layer (SQL Server): Stores, cleans, and processes compliance data, ensuring optimized querying and reporting.
Processing Layer (Power BI & Python/SQL Preprocessing): Handles data transformation, risk assessments, and bias tracking before visualization.
Visualization & Reporting Layer (Power BI Reports & Dashboards): Displays real-time compliance scores, bias detection, and audit logs for regulatory monitoring.

Data Flow Architecture

The system follows a structured data pipeline:

Data Ingestion
- Compliance-related CSV datasets are gathered from multiple sources.
- Data is loaded into SQL Server for storage and processing.

Data Processing
- SQL-based data cleaning and transformation.
- Python scripts for advanced data preprocessing and risk scoring (optional).
- Normalization of tables to improve query performance.

Data Modeling
- Relationship keys are established to connect different compliance-related tables.
- Performance optimization techniques (indexing, partitioning) are implemented to handle large datasets.

Data Visualization & Reporting
- Power BI connects to the SQL Server to extract compliance insights.
- Reports and dashboards are built to display risk assessments, bias tracking, and compliance scores.
- Filters, slicers, and drill-down capabilities enhance data exploration.

Gather CSV data sources

For this demonstration project, we’ll be using a set of prepared CSV files that simulate real-world AI compliance data. While in production these would come from various live systems, our demo files provide a comprehensive dataset for building our compliance monitoring solution. These datasets contain structured compliance-related information, allowing us to build and test the Power BI ReportBook without requiring direct integration with external systems.

Demo Dataset Files

We’ll be working with the following CSV files:

AI_Compliance_Dataset.csv : Core compliance monitoring data
AI_GDPR_Regulatory_Provisions.csv : GDPR-specific regulatory requirements
AI_Model_Interaction_Dataset.csv : Records of AI model interactions and decisions
AI_Performance_Dataset.csv : Model performance metrics and evaluations
AI_Regulatory_Provisions_Dataset.csv : AI Act regulatory requirements
Compliance_Audit_Dataset.csv : Audit logs and findings
Decision_Appeal_Dataset.csv : Appeal records for AI decisions
System_Incident_Dataset.csv : Incident reports and resolutions
User_Feedback_Dataset.csv : User feedback and experience data

The files for the project can be downloaded from the buttons below:

Real-World Data Collection

In a production environment, this data would typically be gathered from various sources:

Automated Sources

AI model monitoring systems
Application logging frameworks
Automated compliance checking tools
Incident management systems
User interaction tracking systems
Performance monitoring tools
Automated audit systems

Manual Input Sources

Compliance officer reviews
Regulatory documentation
Manual audit entries
User appeal submissions
Incident report forms
Stakeholder feedback forms

Integration Considerations

When implementing this system with live data sources, you would need to:

Set up automated data collection pipelines from:
- AI model deployment platforms
- Application servers
- User interaction systems
- Compliance monitoring tools
- Incident management systems
Implement:
- Real-time data validation
- Secure data transfer protocols
- Automated quality checks
- Data transformation rules
- Regular refresh schedules

Data Preparation Steps

Before proceeding with the import, ensure:

All CSV files are:
- UTF-8 encoded
- Free of formatting errors
- Properly structured with consistent headers
- Located in the designated import directory
Verify data quality:
- Check for missing values
- Validate date formats
- Ensure consistent data types
- Verify referential integrity between datasets

With our CSV files prepared, we’ll move on to importing this data into SQL Server for processing and analysis. The variety of our datasets will allow us to create comprehensive compliance monitoring dashboards that cover all aspects of AI regulation requirements.

PHASE[2]
Importing & Preparing the CSV Data

During this phase, we’ll focus on importing our collection of CSV files containing various aspects of AI compliance data – from model performance metrics to regulatory requirements and user feedback. We’ll implement data cleaning procedures, establish proper data structures, and ensure our datasets are optimized for analysis. This groundwork is essential for creating meaningful insights and maintaining accurate compliance tracking.

The quality of our data preparation directly impacts the effectiveness of our compliance monitoring, so we’ll pay careful attention to data validation, schema documentation, and maintaining referential integrity across our datasets. By following a systematic approach to data preparation, we ensure our PowerBI ReportBook will have a reliable foundation for generating compliance insights and risk assessments.

Load CSVs into SQL Server

With our data files cleaned and verified, we’re ready to import them into SQL Server so they can be securely stored, queried, and integrated into our compliance monitoring system. In this step, we’ll create or designate a database to house our tables, then systematically load each CSV file using SQL Server Management Studio (SSMS). By doing so, we’ll be able to leverage SQL queries for data validation, transformation, and the creation of robust compliance dashboards in Power BI.

1. Create a Dedicated Database

Open SQL Server Management Studio and connect to your SQL Server instance on the VM.
Right-click Databases in the Object Explorer and select New Database
Provide a descriptive database name (e.g., Compliance_DB), then click OK to finalize.

2. Use the Import Data Wizard

In SSMS, right-click the new database and go to Tasks → Import Data…
Choose Flat File Source and browse to your first CSV file (e.g., AI_Compliance_Dataset.csv).
Verify the Column Delimiter (commonly comma) and Row Delimiter (commonly {CR}{LF}), then click Next.
Confirm the Destination is set to your newly created database in SQL Server.
In the Select Source Tables and Views step, rename the destination table if desired (e.g., AI_Compliance_Dataset), and ensure the column mappings look correct.
Click Next and then Finish to start the import.

Tip: If you see unexpected characters or alignment issues, check the CSV’s encoding (UTF-8) and delimiters, then re-import as needed.

3. Repeat for All CSV Files

Perform these import steps for each CSV in your dataset, ensuring each table is appropriately named and consistent with your naming conventions. Doing so will keep your compliance-related tables organized and easy to reference in subsequent phases.

4. Verify the Imported Data

After the import, expand the Tables folder under your new database in SSMS.
Right-click a table name and select Select Top 1000 Rows to quickly review the data.
Check that the columns, data types, and row counts match your expectations.
If any anomalies appear (e.g., missing fields, unexpected nulls), address these before moving on—this may involve reformatting the CSV or adjusting import settings.

By completing these steps, you’ll have all your CSV data stored in SQL Server, ready for the data modeling and transformation process ahead.

Perform basic data cleaning

Now that our data is imported into SQL Server, we’ll execute a series of cleaning operations to ensure data quality and consistency across all our compliance-related tables. This step is crucial for reliable analysis and reporting in our PowerBI dashboards.

1. Standardize Text Fields

				
					-- Remove leading/trailing spaces
UPDATE [TableName]
SET [ColumnName] = TRIM([ColumnName])
WHERE [ColumnName] LIKE ' %' OR [ColumnName] LIKE '% '

-- Standardize case for categorical fields
UPDATE [AI_Compliance_Dataset]
SET [ComplianceStatus] = UPPER([ComplianceStatus])
WHERE [ComplianceStatus] IS NOT NULL

2. Handle Missing Values

				
					-- Identify columns with NULL values
SELECT 
    COLUMN_NAME,
    COUNT(*) as NullCount
FROM [Compliance_DB].INFORMATION_SCHEMA.COLUMNS AS cols
LEFT JOIN [TableName] AS data
    ON 1=1
WHERE data.[ColumnName] IS NULL
GROUP BY COLUMN_NAME

3. Fix Date Formats

				
					-- Standardize date formats
UPDATE [TableName]
SET [DateColumn] = TRY_CONVERT(datetime2, [DateColumn])
WHERE ISDATE([DateColumn]) = 1

4. Remove Duplicate Records

				
					-- Identify and remove duplicates based on key fields
WITH DuplicateCTE AS (
    SELECT *,
        ROW_NUMBER() OVER(PARTITION BY [KeyField1], [KeyField2] 
        ORDER BY [TimeStamp] DESC) as RowNum
    FROM [TableName]
)
DELETE FROM DuplicateCTE WHERE RowNum > 1

These cleaning operations should be performed systematically on each imported table, adjusting the specific fields and conditions based on the table’s content and requirements. Document any significant data quality issues encountered and the corrections applied for future reference.

Document each table’s schema

With our datasets imported and cleaned in SQL Server, we now document each table’s schema. This documentation provides a structured overview of the data, including column names, data types, relationships, and usage. Proper schema documentation ensures will help us in the next stage where we refine and re-model the schema to fit the database.

1. AI_Model_Dataset

Description: Stores details on deployed AI models.
Primary Key: id
Foreign Keys: compliance_id
Columns:

Column Name	Data Type	Description
model_id	INT (PK)	Unique feedback entry ID
compliance_id	INT (FK)	compliance table reference
model_name	TEXT	Name of Model
model_type	TEXT	Type of model
creation_date	DATETIME	Date model was created

2. AI_Compliance_Dataset

Description: Stores key compliance monitoring data for AI models, tracking regulatory adherence.
Primary Key: compliance_id
Foreign Keys: model_id (linked to AI_Performance_Dataset)
Columns:

Column Name	Data Type	Description
compliance_id	INT (PK)	Unique ID for each compliance entry
model_id	VARCHAR(50) (FK)	AI model identifier
compliance_status	VARCHAR(20)	Compliance result (e.g., "Compliant", "Non-Compliant")
regulatory_category	VARCHAR(50)	Relevant AI Act category (e.g., "High-Risk AI")
compliance_score	DECIMAL(5,2)	Compliance percentage (0-100)
last_audit_date	DATETIME	Date of last compliance audit
flagged_issues	TEXT	Summary of compliance concerns

3. AI_GDPR_Regulatory_Provisions

Description: Contains GDPR-related AI compliance provisions.
Primary Key: gdpr_id
Foreign Keys: None
Columns:

Column Name	Data Type	Description
gdpr_id	INT (PK)	Unique GDPR provision ID
provision_name	VARCHAR(255)	Name of the GDPR provision
requirement_text	TEXT	Full text of the requirement
risk_level	VARCHAR(20)	Associated risk (e.g., "High", "Medium")

4. AI_Model_Interaction_Dataset

Description: Captures AI model interactions, decisions, and outcomes.
Primary Key: interaction_id
Foreign Keys: model_id (linked to AI_Performance_Dataset)
Columns:

Column Name	Data Type	Description
interaction_id	INT (PK)	Unique ID for each model interaction
model_id	VARCHAR(50) (FK)	AI model identifier
user_id	VARCHAR(50)	Identifier for the user interacting with the model
decision_output	VARCHAR(255)	Model’s decision output
timestamp	DATETIME	Date and time of the interaction

5. AI_Performance_Dataset

Description: Contains AI model performance metrics and evaluation scores.
Primary Key: performance_id
Foreign Keys: model_id (linked to AI_Model_Interaction_Dataset)
Columns:

Column Name	Data Type	Description
performance_id	INT (PK)	Unique ID for performance record
model_id	VARCHAR(50) (FK)	AI model identifier
accuracy_score	DECIMAL(5,2)	Model’s accuracy percentage
bias_score	DECIMAL(5,2)	Bias detection score
performance_date	DATETIME	Date of performance evaluation

6. AI_Regulatory_Provisions_Dataset

Description: Lists specific regulatory requirements under the AI Act.
Primary Key: regulation_id
Foreign Keys: None
Columns:

Column Name	Data Type	Description
regulation_id	INT (PK)	Unique ID for regulation entry
category	VARCHAR(50)	AI risk category (e.g., "High-Risk AI")
requirement_text	TEXT	Description of the regulatory requirement
enforcement_date	DATETIME	Date the regulation takes effect

7. Compliance_Audit_Dataset

Description: Tracks compliance audits conducted on AI models.
Primary Key: audit_id
Foreign Keys: compliance_id (linked to AI_Compliance_Dataset)
Columns:

Column Name	Data Type	Description
audit_id	INT (PK)	Unique audit entry ID
compliance_id	INT (FK)	Reference to compliance dataset
audit_result	VARCHAR(20)	Audit outcome ("Pass", "Fail")
audit_notes	TEXT	Summary of findings
audit_date	DATETIME	Date of audit

8. Decision_Appeal_Dataset

Description: Logs user appeals against AI decisions.
Primary Key: appeal_id
Foreign Keys: interaction_id (linked to AI_Model_Interaction_Dataset)
Columns:

Column Name	Data Type	Description
appeal_id	INT (PK)	Unique appeal record ID
interaction_id	INT (FK)	Reference to model interaction
appeal_reason	TEXT	User’s reason for appeal
resolution_status	VARCHAR(50)	Appeal resolution status ("Pending", "Approved", "Rejected")
resolution_date	DATETIME	Date of resolution (if applicable)

9. System_Incident_Dataset

Description: Records system incidents affecting AI compliance.
Primary Key: incident_id
Foreign Keys: model_id (linked to AI_Performance_Dataset)
Columns:

Column Name	Data Type	Description
incident_id	INT (PK)	Unique system incident ID
model_id	VARCHAR(50) (FK)	AI model associated with the incident
incident_type	VARCHAR(50)	Category of incident ("Bias Detected", "Performance Drop")
incident_description	TEXT	Description of the issue
resolution_status	VARCHAR(50)	Incident resolution status
incident_date	DATETIME	Date the incident was reported

10. User_Feedback_Dataset

Description: Stores feedback from users interacting with AI models.
Primary Key: feedback_id
Foreign Keys: model_id (linked to AI_Model_Interaction_Dataset)
Columns:

Column Name	Data Type	Description
feedback_id	INT (PK)	Unique feedback entry ID
model_id	VARCHAR(50) (FK)	AI model referenced
user_id	VARCHAR(50)	Identifier for feedback provider
feedback_text	TEXT	User’s feedback message
feedback_date	DATETIME	Date feedback was submitted

PHASE[3]
Data Schema Modelling

Now that we have our data imported and cleaned, we’ll focus on structuring it optimally for compliance monitoring and analysis. This phase involves creating relationships between our various compliance datasets, transforming data where needed, and establishing a robust schema that supports efficient querying and visualization.

During this phase, we’ll organize our compliance data using established dimensional modeling principles, ensuring that metrics, risk assessments, and regulatory requirements are properly linked. We’ll create appropriate relationships between fact tables (like model interactions and compliance events) and dimension tables (such as regulatory provisions and model metadata), enabling powerful cross-table analysis and comprehensive compliance tracking.

The transformations we implement will focus on:

Creating consistent keys across compliance-related tables
Establishing proper relationships between regulatory requirements and compliance data
Optimizing data structures for performance in Power BI
Implementing calculated fields and measures for compliance scoring
Ensuring data models support both real-time monitoring and historical analysis

Identify or create relationship keys

Establishing relationship keys is essential for creating a structured and optimized data model that allows seamless integration between datasets, ensuring accurate compliance tracking, bias analysis, and risk assessment. In this section, we define the primary keys (PKs) and foreign keys (FKs) needed to connect different tables while maintaining data integrity.

To create an effective data model, we need to identify unique keys in each dataset that establish relationships between compliance data, AI performance, regulatory provisions, and user interactions.

NOTE TO SELF: we have renamed tables (see capitalized in tables)

NOTE TO SELF: we transform model_id etc in tables to be id etc … then on t-sql procedure with ingestion we transform … bulk insert

NOTE TO SELF: The need for one-to-many tables f.ex. in
GDPR_PROVISIONS id → aiact_provision_id AI_COMPLIANCE

AIACT_PROVISIONS id → gdpr_provision_id AI_COMPLIANCE

The following Tables Summarize these relationships for our data.

Primary / Foreign Key Overview

Dataset	Primary Key (PK)	Foreign Keys (FKs) (Links to other datasets)
GDPR_PROVISIONS	Provision ID	None (lookup table)
MODEL_INTERACTION	Interaction ID	AI Model → AI models in other datasets
PERFORMANCE	Performance ID	AI Model ID
AIA_PROVISIONS	Provision ID	None (lookup table)
AUDIT	Audit ID	AI Model ID → AI models in other datasets
DECISION_APPEAL	Appeal ID	AI Model ID, User ID
AI_COMPLIANCE	AI Model ID	Regulatory Framework, Compliance Status
SYSTEM_INCIDENT	Incident ID	AI Model ID
USER_FEEDBACK	Feedback ID	AI Model ID, User ID

Relationship Mapping Table

Source Table	Relationship	Destination Table
AI_MODEL	id → model_id	MODEL_INTERACTION
AI_MODEL	id → model_id	AUDIT
AI_MODEL	id → model_id	SYSTEM_INCIDENT
AI_MODEL	id → model_id	USER_FEEDBACK
AI_COMPLIANCE	id → compliance_id	AI_MODEL
GDPR_PROVISIONS	id → aiact_provision_id	AI_COMPLIANCE
AIACT_PROVISIONS	id → gdpr_provision_id	AI_COMPLIANCE
MODEL_INTERACTION	id → interaction_id	DECISION_APPEAL

Relationship Mapping Chart

organization ID, model ID, date

Normalize or flatten tables

snowflake schema, data vault v2

PHASE[4]
ETL (Extract, Transform, Load) & Procedural T-Sql

fuzzy matching, bridging tables

Finalize database schema

fact tables, dimension tables, or separate tables

T-SQL

organization ID, model ID, date

Optional Python/SQL preprocessing

organization ID, model ID, date

PHASE[4]
Set up version Control

Git version control repository for t-sql scripts and code

direct query or import mode

Handling data structure migrations

direct query or import mode

PHASE[5]
Building the PowerBI ReportBook

This guide provides step-by-step instructions for creating a PowerBI studies management visualization (viz). The viz will automatically update itself if the data inside any of the excel columns are changed.

Here is also a Youtube video that goes through the same step-by-step guide that is presented in written form here.

Under development / More coming soon

Connect Power BI to SQL Server

direct query or import mode

Create core visuals

compliance trends, risk heatmaps, bias metrics, user appeals

Set up data refresh

manual or scheduled updates

Apply filters & slicers

date ranges, model selection, compliance tiers

PHASE[6]
Compliance & Risk Analysis

Here is also a Youtube video that goes through the same step-by-step guide that is presented in written form here.

Under development / More coming soon

Automated risk scoring

leveraging data from “risk_assessment.csv”

Bias detection dashboards

group fairness metrics, user feedback

User appeals & escalation tracking

incident logs, resolution times

Correlate compliance scores with risk & performance

integrated visuals

PHASE[7]
Dashboard Publishing & Sharing

Here is also a Youtube video that goes through the same step-by-step guide that is presented in written form here.

Under development / More coming soon

Publish to Power BI Service

integrated visuals

Set up gateways

(if on-prem SQL) and refresh schedules

Enable alerts & subscriptions

compliance drops, risk surges

row-level security, user permissions

compliance drops, risk surges

PHASE[8]
Maintenance & Automation

Here is also a Youtube video that goes through the same step-by-step guide that is presented in written form here.

Under development / More coming soon

DevOps & version control

GitHub Actions, Azure DevOps pipelines

Monitor usage & performance

Power BI usage metrics, SQL Server logs

Plan future enhancements

additional data sources, advanced analytics

Stay updated with EU AI Act changes

amendments, new regulations

0 Comments

Inline Feedbacks

View all comments

SHARE

TABLE OF CONTENTS

CONTENTS

PHASE[1] Project Setup & Requirements

Define the overall objectives

Install required tools

Plan the architecture

Gather CSV data sources

PHASE[2] Importing & Preparing the CSV Data

Load CSVs into SQL Server

Perform basic data cleaning

Document each table’s schema

PHASE[3] Data Schema Modelling

Identify or create relationship keys

Normalize or flatten tables

PHASE[4] ETL (Extract, Transform, Load) & Procedural T-Sql

Finalize database schema

T-SQL

Optional Python/SQL preprocessing

PHASE[4] Set up version Control

Git version control repository for t-sql scripts and code

Handling data structure migrations

PHASE[5] Building the PowerBI ReportBook

Connect Power BI to SQL Server

Create core visuals

Set up data refresh

Apply filters & slicers

PHASE[6] Compliance & Risk Analysis

Automated risk scoring

Bias detection dashboards

User appeals & escalation tracking

Correlate compliance scores with risk & performance

PHASE[7] Dashboard Publishing & Sharing

Publish to Power BI Service

Set up gateways

Enable alerts & subscriptions

row-level security, user permissions

PHASE[8] Maintenance & Automation

DevOps & version control

Monitor usage & performance

Plan future enhancements

Stay updated with EU AI Act changes

SHARE

PHASE[1]
Project Setup & Requirements

PHASE[2]
Importing & Preparing the CSV Data

PHASE[3]
Data Schema Modelling

PHASE[4]
ETL (Extract, Transform, Load) & Procedural T-Sql

PHASE[4]
Set up version Control

PHASE[5]
Building the PowerBI ReportBook

PHASE[6]
Compliance & Risk Analysis

PHASE[7]
Dashboard Publishing & Sharing

PHASE[8]
Maintenance & Automation