Our services
Four building blocks, one line. Responsible AI in documented practice.
We don’t sell tools or license fees. We sell a method that takes your AI practice from a non-binding statement to a provable, auditable state — step by step, at a pace that fits your organization and the market access you’re working toward.
The logic
From understanding, through testing, to making it last.
The four building blocks build on one another. Each can be booked on its own, but their full effect comes from running them as a sequence.
S4 · Understand
Workshops
Close the knowledge gap in leadership and engineering — before external advice costs money that could be built in-house.
S1 · Locate
Readiness Assessment
Inventory, risk classification, role determination, compliance gaps. The foundation everything else is built on.
S2 · Test
Ethics & Bias Audit
Deep technical audit of specific systems: fairness, explainability, data quality — reproducibly documented.
S3 · Sustain
Continuous Monitoring
Drift detection, bias alerting, and post-market monitoring under Art. 72 — AI governance is a state, not a project.
The building blocks in detail
Four services, each bookable on its own.
EU AI Act Readiness Assessment
A systematic stocktake: which AI systems do you use or ship, which risk class do they fall into under the EU AI Act, which role do you hold, and which compliance gaps must you close by which deadline? The result is a prioritized roadmap — so you know where you stand and in what order to act.
What you get
- Complete AI system inventory (Python dashboard)
- Risk classification matrix under the EU AI Act
- Provider / deployer role determination
- Compliance gap analysis per system
- Strategic 12–24 month roadmap
- Executive presentation for board / leadership
Makes sense if: you run between 2 and 15 AI systems, you’re unsure which fall under the Act, and you need clarity for budget and resource planning or for answering EU customers.
AI Ethics Audit & Bias Testing
Where the Readiness Assessment asks “where do we stand?”, the audit asks “how do our models actually behave?”. We measure fairness quantitatively with established statistical methods — statistical parity, equal error rates, equalized odds — and explain where and why your models decide systematically differently.
What you get
- Bias detection report (executive + technical)
- Fairness metrics across protected attributes
- Intersectional bias analysis
- Explainability report (SHAP-based)
- Data quality assessment
- Fairness monitoring dashboard (Python / Streamlit)
Makes sense if: you operate concrete high-risk systems (recruiting, credit scoring, recommendation, automated decisions) and either face a conformity step or must document external credibility for EU buyers.
Continuous AI Monitoring
Bias isn’t a static property. It shifts over time through data drift, user behavior, and feedback loops — we call it bias creep. Test once and stop watching, and you’re trusting a false sense of safety. Continuous Monitoring meets the post-market monitoring duty under Art. 72 and gives you the evidence base for quarterly board updates.
What you get
- Drift detection (Alibi Detect / Evidently AI)
- Temporal fairness metrics with alert thresholds
- Data quality pipeline (Great Expectations)
- Monitoring dashboard for compliance owners
- Quarterly compliance reviews with report
- Escalation paths and accountability matrix
Makes sense if: your AI systems are in production, have high-risk characteristics or ESG reporting relevance — and you need regular documentation for investors, rating agencies, or your board.
Responsible AI Workshops
Advice costs less when you know what you’re talking about. Three formats build the internal capability you’ll still need after the best audit: Executive Education for leadership, Technical Team Training for data scientists and engineers, and Ethics Framework Development for cross-functional teams.
What you get
- Executive Education (1 day, 8–15 participants)
- Technical Team Training (1–2 days, 10–20 participants)
- Ethics Framework Development (2 days, 12–25 participants)
- Industry-specific case studies
- Python tools and templates to take away
- Follow-up 30-min office hours included
Makes sense if: you want to build internal confidence before releasing external budget — or, after an assessment, to implement identified measures with your own teams.
Three packages
How we bundle the building blocks.
The three packages are the most common combinations we run with clients — staged by maturity and scope. They include bundle discounts versus booking the building blocks individually.
| Component | Foundation | Implementation | Transformation |
|---|---|---|---|
| EU AI Act Readiness Assessment | ✓ | ✓ | ✓ |
| Strategic roadmap & quick wins | ✓ | ✓ | ✓ |
| Bias Audit (focus system) | — | ✓ | ✓ |
| Workshops (Executive + Technical) | — | ✓ | ✓ |
| Continuous Monitoring setup | — | ✓ | ✓ |
| Bias audit portfolio (all systems) | — | — | ✓ |
| Monitoring retainer (12 months) | — | — | ✓ |
| Quarterly compliance reviews | — | — | ✓ |
| Price | from €18,000 | from €35,000 | from €60,000 |
| Duration | 4–6 weeks | 3 months | 6+ months |
All packages are entry prices and are finalized based on your AI portfolio (number of systems, complexity, risk class). Individual building blocks are also bookable outside the packages. Prices are net of VAT.
We don’t sell hours. We sell a state — one you can document and prove after our work is done.
— Dr. Valentin José Mayr · Founder
Common questions
What clients ask before the first call.
Do we have to start with the Readiness Assessment — or can we book an audit directly?
You can book an audit directly if you know exactly which system needs testing and why. If that clarity isn’t there — “we run ten AI systems and don’t know which are critical” — we start with the assessment. Otherwise you risk auditing the wrong system.
What’s the difference between an audit and continuous monitoring?
An audit is a snapshot: we measure fairness, explainability, and data quality at a defined point in time and hand over a report. Six months later the model may behave differently as input data shifts.
Continuous Monitoring watches those shifts automatically and alerts when thresholds are crossed. Both matter — audit for point-in-time proof, monitoring for the post-market obligation under Art. 72.
How does this fit with our existing GDPR compliance?
GDPR and the EU AI Act overlap in two main places: automated decision-making (Art. 22 GDPR ↔ Art. 6 + 26 EU AI Act) and special categories of data (Art. 9 GDPR ↔ Art. 10 EU AI Act). Elsewhere they’re distinct regimes with their own duties.
We work explicitly from your GDPR documentation as a starting point and identify what’s reusable for the AI Act and what must be created. It’s usually less new work than feared.
We’re not in the EU. Why would we need this?
Because the Act reaches you anyway if your AI is placed on the EU market or its outputs are used in the EU — and because your EU customers will ask for evidence in procurement. Our scope page walks through exactly when you’re covered.
Your next step