Zum Inhalt springen
Nahaufnahme von Händen auf einer Laptop-Tastatur, auf dem Bildschirm Analytics-Daten und Code

Built with AI. Audited independently.

You’re already using software components built with AI tools and want to make sure those solutions still meet the highest standards for quality and security? Together with our partner TÜV TRUST IT, we offer exactly that: the “TÜV Trusted Development Certificate”, now extended to also assess how AI tools are applied.

Our own website was our first test case: we migrated it to a headless architecture with Next.js, Storyblok and Vercel, working with AI tools throughout. Code, content structure and tests all came out of that process. The audit happened on the live system, the one that’s online today.

Speed is not proof

AI is changing how software gets built. What used to take weeks now takes days: an AI agent reads a requirement, builds the component, writes the test for it and documents the decision. That’s how we work, every day.

That raises a new question, and it gets louder with every project: who actually checked this? As long as people wrote every line themselves, the answer was obvious. There were reviews, experience, and someone who stood behind their code. When a large share of the work is done by AI, that answer has to be made explicit.

That’s why we decided early on not to leave the proof to trust. A claim that rests on trust alone is just a claim with a good reputation.

Our review chain

Three layers, each building on the last

Each layer checks something the one before it can’t.

Build, check, have it checked

  1. Layer 1

    Built with AI

    AI agents implement components, content models and pages. Every change is traceable, every architecture decision documented. A person signs off on every step.

  2. Layer 2

    Checked by ourselves

    Specialized review agents check every change against our standards: accessibility, load time, brand presentation and security. Automated tests run alongside every change.

  3. Layer 3

    Audited independently

    TÜV TRUST IT GmbH reviewed how we work with AI tools, using our own website as the subject: through interviews, document and source code review, and a penetration test. The result was entirely up to the auditor.

What was audited

We check

AI in development

How AI tools and agents are embedded in development: governance · control · and the measures against the risks from the OWASP Top 10 for LLM applications.

Development process

From requirements · architecture and planning through testing · secure coding and source control · to change management · maintenance and operations.

Data protection and components

Information and data protection · plus the tools and third-party components in use.

Penetration test

A penetration test of our website as the reference environment.

The result

Used responsibly, implemented securely

TÜV TRUST IT confirmed the responsible use of AI agents in our development process and rated the security implementation of our website as convincing.

Certificate confirmed

The audit doubled as the scheduled review of our Trusted Development certificate, which we’ve held since 2024. The certificate is confirmed.

And one finding that stays with us

Security reviews shouldn’t be carried out solely by the same AI agent or language model that was already used during development. That was our thesis. The independent audit reaches the same conclusion.

SHOPMACHER hat in diesem Projekt eindrucksvoll gezeigt, wie sich moderne KI-gestützte Entwicklungsansätze erfolgreich in die Praxis überführen lassen. Besonders überzeugt haben uns die strukturierte Umsetzung, die Offenheit für eine unabhängige Überprüfung und der hohe Anspruch an die Qualität und Sicherheit der entwickelten Lösung.

André Zingsheim

Principal Consultant, TÜV TRUST IT GmbH

MACH Alliance

Our position

Independence is not a formality in AI-assisted development

AI agents are very good at checking against rules we’ve given them. Rules nobody thought of, they won’t find. That’s exactly where an external reviewer comes in, with their own catalogue and their own perspective.

The regulatory environment points the same way. Germany’s Barrierefreiheitsstärkungsgesetz (Accessibility Strengthening Act) has applied since June 2025, the EU AI Act is phasing in, and anyone building an AI management system under ISO/IEC 42001 needs solid evidence. We aligned our own management system to it before recommending it to anyone else.

What you’re probably asking now

Let’s talk about your project

In an intro call, we’ll figure out which of the three paths fits you and what the process looks like.

Book an intro call
Manuel Ludvigsen-Diekmann

Manuel Ludvigsen-Diekmann

CTO

Built with AI. Audited independently. | SHOPMACHER