Our approach: Systematic, practical, implementable

The waveImpact EU AI Act Readiness Assessment is not a theoretical compliance check, but a practical assessment of your AI landscape with a concrete implementation roadmap. We combine regulatory expertise with technical depth and provide you not only with a diagnosis, but also with a solution.

Our assessment is based on three pillars:

1. Complete inventory
We systematically record all AI systems in your company—from productive applications to pilot projects. We use structured interviews, document analysis, and technical inspections. The result: a complete, categorized AI system registry.

2. Compliance gap analysis
For each identified system, we check the applicability of the EU AI Act and perform a detailed gap analysis. We evaluate not only the obvious requirements, but also the technical details: Is there bias? Is the system explainable? Is the data quality sufficiently documented?

3. Prioritized action plan
The most important question: What do you need to do and when? We develop an implementation roadmap prioritized according to risk and effort, with realistic timelines and budget estimates. This includes quick wins, must-haves, and nice-to-haves.

How we work: 6-8 weeks to a complete compliance overview

Phase 1: Kickoff & Discovery (Week 1)

🚀 Start

Activities:

• Kickoff workshop (half day, on-site or remote)
• Stakeholder interviews (C-level, legal, IT, data science)
• Document review (existing AI documentation)
• Clarify access to systems and documentation
• Detailed project setup

Output:

• Project plan with timeline
• stakeholder matrix
• Initial system list (draft)

Your time commitment: 1 day (workshop + interviews)

Phase 2: System inventory (weeks 1-3)

📋 Recording

Activities:

• Structured recording of all AI systems
• Technical interviews with development teams
• Architecture review (system design, data flows)
• use case documentation
• business impact assessment
• Responsibility mapping

Output:

• Complete AI System Inventory
• Initial categorization by type and risk
• Technical profiles for each system

Your time commitment: 2-3 days (team interviews spread out)

Phase 3: Risk classification & compliance check (weeks 3–4)

⚖️ Rating

Activities:

• EU AI Act Applicability Assessment per System
• Risk classification according to Annex III
• Article-by-article compliance check (Articles 9–15, 72)
• GPAI classification
• Prohibited practices check
• Integration with GDPR compliance

Output:

• Risk classification report
• Compliance status matrix (target vs. actual)
• List of affected items per system

Your effort: 0.5 days (queries)

Phase 4: Technical validation (weeks 4-6)

🔬 Testing

Activities:

• Bias & Fairness Testing (High-Risk Systems)
• Explainability Analysis
• Data Quality Assessment
• Robustness Testing (Basic)
• documentation audit
• Technical Risk Assessment

Use of technology:

• IBM AI Fairness 360
• SHAP (Explainability)
• Great Expectations (Data Quality)
• Custom Analysis Scripts

Output:

• Technical compliance reports per system
• Bias analyses with visualizations
• Explainability dashboards
• Data Quality Scorecards

Your effort: 1-2 days (data access, technical queries)

Phase 6: Presentation & Handover (Weeks 7-8)

🎯 Completion

Activities:

• Executive Summary Finalization
• Management presentation (on-site recommended)
• Detailed walk-through of all deliverables
• Q&A session with stakeholders
• Handover workshop (next steps)
• Define follow-up plan

Output:

• Final documentation (all deliverables)
• management presentation
• handover protocol
• Optional: Follow-up offer (implementation support)

Your time commitment: 1 day (presentations + workshop)

>> Total effort for your team: 5-8 working days spread over 6-8 weeks

Who is this assessment suitable for?

Other industries:

• Public administration (e-government, social services)
• Retail (personalization, pricing)
• Energy & Utilities (Smart Grid, Demand Forecasting)
• Logistics (route optimization, warehouse automation)
• Telecommunications (network optimization, customer service)

Our technical expertise: Tools that deliver results

Bias and Fairness Testing

🔍 IBM AI Fairness 360

The leading open-source framework for fairness analysis. We use it for:

• 70+ fairness metrics (disparate impact, equal opportunity, etc.)
• Pre-processing, in-processing, post-processing mitigation
• Group Fairness & Individual Fairness Tests
• Intersectional analyses (e.g., gender + age combined)

Use: Credit scoring, HR systems, access control

🔍 Microsoft Fairlearn

For constraint-based fairness optimization:

• Simpler client communication than AIF360
• Trade-off analyses: Fairness vs. Accuracy
• Grid search for fair hyperparameters
• Integration with scikit-learn

Use: Smaller models, prototyping, client workshops

Explainability & Transparency

💡 SHAP (SHapley Additive exPlanations)

The gold standard for model explainability:

• Feature importance at prediction level
• Global & local declarations
• Visualizations for technical and non-technical stakeholders
• Model-agnostic (works for almost all ML models)

German market requirement: “Why did the system make this decision?”
SHAP provides the answer.

💡 LIME (Local Interpretable Model-agnostic Explanations)

Backup tool for specific use cases:

• Faster calculations than SHAP
• Text & Image Explainability
• Model-agnostic

Application: NLP systems, computer vision, complex ensembles

Data Governance & Quality

📊 Great Expectations

Data Quality Framework for AI Compliance:

• Automated data validation
• Expectation Suites (rules for data quality)
• Data Profiling & Documentation
• Audit trails (GDPR & AI Act compliant)

EU AI Act Art. 10 Requirement: “Training, validation, and testing data sets shall be relevant, sufficiently representative, and to the best extent possible, free of errors and complete.”

Great Expectations makes this requirement measurable and documentable.

Monitoring & Drift Detection

📈 Alibi Detect (for advanced assessments)

If a need for monitoring is identified:

• Data drift detection
• Model drift detection
• Outlier detection
• Adversarial detection

Application: Post-market monitoring plan development

methodology

Our assessment follows established frameworks:

• EU AI Act Conformity Assessment (Art. 43)
• ISO/IEC 42001:2023 (AI Management System)
• NIST AI Risk Management Framework
• IEEE 7000 Series (Ethically Aligned Design)
• BSI AI Cloud Service Compliance Criteria Catalog (AIC4)

Combined with proprietary checklists for German compliance culture.

Transparent pricing

Add-ons (optional):

• Additional technical validation per system: $3,000 – $8,000
• On-site workshops (additional): $1,500/day + expenses
• Express service (50% time reduction): +30% surcharge
• Ongoing support retainers: Starting at €2,500/month

Payment terms:

• 30% at project start
• 40% upon interim presentation (end of phase 4)
• 30% upon final acceptance

Invoicing in accordance with German tax law (VAT deductible)

FREQUENTLY ASKED QUESTIONS (FAQ)