AI AT
SHOPMACHER
How Shopmacher work with artificial intelligence
AI as a tool, not a replacement
Artificial intelligence is changing the way digital products are built. Shopmacher use AI deliberately to develop commerce platforms faster, better and in a more data-driven way. One principle is clear: AI supports, people decide. This overview shows what that means in practice, what everyday work looks like and which structures ensure AI is used responsibly.
For Shopmacher, AI is a supporting technology. It helps with analysis, speeds up development processes and generates drafts for code, content and test cases. The professional responsibility always stays with people. No AI-generated result leaves the house without being reviewed and approved by a qualified person. That applies to customer platforms just as much as to internal concepts, proposals and documentation.
This approach is a conscious choice: AI systems work probabilistically, they deliver answers that are highly likely to be correct. In many contexts that is enough. When it comes to contracts, technical specifications or customer communication, however, likely is not enough. That is why Shopmacher review every AI-generated output professionally before it is used further.
Six core principles for using AI
The way Shopmacher handle AI is based on six core principles, anchored in a company-wide AI policy.
The first principle is human responsibility. AI makes no decisions on its own when it comes to customer questions, business decisions or legal matters. Every AI application has a responsible person who reviews and approves the result.
The second principle is transparency. Shopmacher document which AI is used, with which inputs and with which results. Customers are informed when AI has made a significant contribution. Documents created largely with AI carry a corresponding note.
The third principle concerns security and stability. No AI system goes into production untested. Before every use, a risk assessment takes place, changes are controlled and failure scenarios are considered.
Data protection and confidentiality form the fourth principle. The existing data protection policies apply in full. Personal data is only processed within the permitted scope. Sensitive company or customer data is not fed into external AI systems in an uncontrolled way.
The fifth principle is fairness and quality. AI systems are checked for discriminatory effects, and results that are relevant to customers undergo a professional quality check.
The sixth principle follows a whitelist approach. Only approved AI tools from a central AI register may be used for business purposes. Every new tool goes through an assessment by those responsible for AI before it is put to use.
AI competence: the Shopmacher AI license
Shopmacher do not just deploy technology, they also invest deliberately in the competence of their team. Every employee completes a mandatory AI license. This training program conveys a solid basic understanding of AI and generative systems, highlights opportunities and limits, and trains responsible use in everyday work.
The AI license is validated through an exam with the International Certification Organization (ICO) and is a prerequisite for using approved AI tools in the company. Depending on the role, there are different formats: a compact e-learning version for all employees and an extended version with deeper governance and compliance knowledge for those responsible for AI and for leaders.
With this, Shopmacher already meet the requirements of Article 4 of the EU AI Act, which since February 2025 has demanded a sufficient level of AI competence from all providers and operators.
Governance: guardrails with room to move
AI governance at Shopmacher follows one guiding idea: clear rules that provide orientation without slowing down innovation. That means working with AI in a structured way that still leaves room for creativity and experimentation.
At its center is an AI management system aligned with ISO/IEC 42001:2023. It covers the six core principles, a central AI register with an approval process, the mandatory AI license, and structured feedback and incident management. Anomalies, errors and suggestions for improvement are recorded through defined reporting channels, assessed and turned into concrete measures.
Responsible for AI governance is CTO Manuel Ludvigsen-Diekmann, who as a triple ICO-certified AI expert per ISO/IEC 42001:2023 (AIMS Officer, AI Compliance Expert, AI Manager) owns the strategic development and operational implementation.
AI hygiene: deciding deliberately when AI is not the right choice
Handling AI responsibly also means knowing when it is better not to use it. Shopmacher have defined five hygiene rules for this that serve as orientation in everyday work.
Personal communication is written by hand. Birthday wishes, thank-yous and personal feedback come from a person, not from a machine. Human relationships deserve real words.
AI output is not passed on unchecked. Anything that goes to colleagues, customers or partners has to be understood in substance and reworked in your own words. Superficially generated texts that others first have to laboriously interpret are known at Shopmacher as "workslop" and are expressly not wanted.
Raw AI output is not forwarded to others unasked. Everyone can query a language model themselves and needs no middleman for it. Checking and refining your own thoughts with the help of AI is fine, but "I asked ChatGPT" is not a contribution to a professional discussion.
When the thinking process itself is valuable, AI is left out. Strategic considerations, conceptual work and learning processes benefit from people going through them themselves. The insight gained from your own thinking is worth more than the time saved by delegation.
And when an answer absolutely has to be correct, it is verified by experts. Language models deliver likely answers, not guaranteed ones. For legal statements, contract details or safety-relevant decisions, the responsibility lies with people.
The five etiquette rules
For centuries a simple convention held: whoever writes a longer text has put in effort. Generative AI changes that fundamentally, because pages-long messages can be produced in seconds. For the first time in history, a text can demand more effort from the person receiving it than from the person sending it.
This asymmetry should be handled consciously. People who matter to you do not deserve an AI-generated text. In human relationships, efficiency has no place.
For us that means: personal messages to colleagues, customers or partners are written by hand. Birthday wishes, thank-yous, feedback and personal assessments come from a person, not from a machine.
The term "workslop" (coined in 2025 by researchers from Stanford and the BetterUp Labs) describes a growing problem: employees send each other AI-generated reports, presentations and emails. At first glance these results look structured and eloquent; on closer inspection they turn out to be superficial and often useless.
The problem is the shifted burden: the recipients have to laboriously interpret the AI-generated content and check it for substance. The productivity of the whole team suffers. AI may support your own work, but the result passed on to others must be understood in substance by the sender, reviewed and reworked in their own words.
"I asked ChatGPT, and this is the answer" is not an acceptable contribution to a professional discussion. Passing on AI output unchecked mainly shows that the person did not bother to think for themselves. It is comparable to saying: "I copied the first Google result without reading it."
The rule of thumb is: checking and refining your own thoughts with the help of AI is fine. Forwarding raw AI output to others who did not ask for it is not. Everyone is capable of asking a language model themselves and needs no middleman for it.
The research is now clear: when you outsource thinking to language models, different and fewer areas of the brain are active. Summarizing papers, structuring thoughts, untangling meeting notes, creating research dossiers, AI can support all of it. But less sticks when you do not go through the thinking process yourself.
So the rule is: when there is enough time and the topic is important, you should skip using language models. The insight gained from your own thinking process is worth more than the time saved by delegating to a machine. This applies in particular to strategic considerations, conceptual work and learning processes.
Modern language models confabulate less often than their predecessors, but they remain probabilistic systems that give answers which are likely correct. For many use cases that is sufficient, for some it is catastrophic.
An answer that is 98 percent correct is not enough in many professional contexts. That is especially true for legal statements, medical information, contract details, financial calculations and safety-relevant decisions. Language models are only suitable for such scenarios if they are contained by RAG systems with verified sources and complemented by human expertise.
For us that concretely means: AI-generated content in proposals, contracts, technical specifications and customer communication must always be reviewed professionally. The responsibility for correctness lies with people, not with the machine.
A summary for everyday work
The five rules can be condensed into a simple checklist. Before using AI, ask yourself: is this about a personal relationship? Then write it yourself. Will the result be passed on to someone? Then understand it in substance and put it in your own words. Did the other person ask for AI output? If not, formulate your own thoughts. Is the thinking process itself valuable? Then think it through yourself. Does the answer absolutely have to be correct? Then verify it or skip AI.
Why this matters
For Shopmacher customers this approach means: you work with a partner who consistently taps AI's potential without giving up quality or responsibility. Every result is reviewed professionally, every use documented and every tool assessed in advance. That builds trust and transparency.
For talent looking for a technologically advanced environment, it shows: at Shopmacher AI is not a buzzword but lived everyday practice with clear structures. Those who work here are trained, allowed to experiment and actively shape how AI is used in e-commerce development.
Created with the help of AI, reviewed and approved by Manuel Ludvigsen-Diekmann, CTO.
Interested in working together?
Shopmacher combine AI competence with years of commerce experience. Let us find out together how AI can move your digital business forward.
You might also be interested in
Values-based Guidelines
What we stand for and how we work.
Careers at Shopmacher
Open positions and insights into how we work.
The Guesswork Trap: Why digital projects fail before they begin
Most digital projects fail not because of the chosen technology, but because of a dangerous mix of unclear starting points, vague requirements and unclear responsibilities.