Services
Put AI to work in your company. We check it before it goes live.
We help small and medium-sized companies in Luxembourg use AI in their daily work. Pick one of the three services below. Do you build AI for your own clients? Our services for agencies are further down the page.
What we do
Pick the engagement that fits where you are
Safe AI
AI governance: we check every system before it goes live
Most AI problems start on launch day. A chatbot forgets to say it is an AI. Client data ends up in a tool nobody checked. Or the AI vendor updates its model, and the answers change without warning.
Before any system we build goes live, we run it through a short checklist we call the release gate. It has six checks and ends in one decision: launch it, or fix it first.
The six checks
Role and risk level
The EU AI Act gives different duties to the company that builds an AI (the provider) and the company that uses it (the deployer). We work out who is who, and which risk level applies: banned, high-risk, transparency duties only, or minimal.
Personal data
We trace where personal data goes and who handles it, including the AI vendor and where it stores the data. We also check whether you need a data protection impact assessment (DPIA).
Testing
We test on real cases. We run each test about ten times and score every answer, because AI answers vary. We also try to trick the system into doing something unsafe.
People stay in charge
People are told they are talking to an AI, and they can always reach a person. The AI asks a person before it sends, pays or deletes anything. And there is an off switch.
Logs and model updates
We decide what to record and for how long. We lock the system to one model version, and when the vendor brings out a new one, we test again.
Vendor terms and handover
We read the AI vendor's terms and put the right clauses in the client contract. When the project ends, you keep all the documents.
Checked by TealSpark
An outside check before you launch
Already have an AI system, built by us or by someone else? We can run the same six checks on it. You get a short report on what passed and what to fix, and a dated note you can show your clients.
This is a practical check, not a legal opinion or a certification, and the report says so.
Why now? Since 2 August 2026, a chatbot or voice agent must tell people it is an AI. Companies that use AI must also take steps to help their staff understand it. Under the AI Omnibus, which came into force in July 2026, most high-risk duties start on 2 December 2027.
For AI agencies
Working with agencies that build AI
Do you build chatbots or automations for clients? Many agencies do good work, but they don't have a clear view of the real risks, or of how to assess them. And their clients ask about those risks more and more. So we cooperate with agencies like yours. We review the processes you run for clients, help you build stronger AI governance into them, and teach your team what to look for.
AI governance for agencies
We review how you work with clients and set up the six checks inside your agency, so your team runs them on every project.
- An overview of the processes you run for clients, with what to improve
- A half-day session with your team, on one of your own projects
- Templates for the six checks, client contracts and data processing
Technology risk advice for agencies
We look at the risks in what you ship and help you decide what to fix first.
- Risks from the AI vendor, the hosting and where the data goes
- Attacks hidden in prompts, surprise model changes and a plan for when something goes wrong
- A “checked by TealSpark” review of your most important client systems
Questions
What people ask about AI governance and the AI Act
AI governance means a few simple rules for how your company chooses, tests, uses and watches its AI tools. For a small company, it can fit on one page: a list of your AI tools, who is responsible for each one, what data each one touches, and a check before anything new goes live.
Very likely, if you use or offer AI. The AI Act covers companies that build AI systems (providers) and companies that use them at work (deployers). Most small companies are deployers. Today their main duties are to tell people when they are talking to an AI, and to help their staff understand AI.
Bill 8476, filed in December 2024, would make the CNPD the main authority for the AI Act. The CNPD is Luxembourg's data protection authority. Sector regulators keep their own areas. The CNPD also runs a regulatory sandbox, where companies can test an AI system with its guidance.
It must tell people they are talking to an AI, unless that is obvious. It should also let them reach a person and handle personal data under GDPR. And it should be tested and watched, so its answers stay right when the model changes.
It should check who is the provider and who is the deployer, the risk level and where personal data goes. It should also check how the system was tested, how a person can step in, what is logged and what the AI vendor's terms say. Our release gate puts all of this into six checks on one page.
No. It is a practical check by an outside team. You get a report and a dated note that show what was checked. It is not a legal opinion or a certification.
Yes. We start with a list of the AI tools you already use. Then we run the six checks on the ones that matter most and fix the gaps with you, without stopping the work that already runs.
The method
Every engagement runs on MODE
Map. Orchestrate. Deliver. Evolve. Four steps that take you from questions to running systems — and keep you evolving after that.
Explore the MODE methodReady to switch into future-proof MODE?
A 30-minute strategy call is enough to know where AI can move your business first.
