The EU AI Act

The EU AI Act is in force. The work is the same, the urgency is new.

Every company that uses AI at work must now be able to show that its people are trained, that its assistants say they are AI, and that AI-made content is marked. If the AI is built properly in the first place, none of this needs a separate compliance project. That is how we build.

Practical guidance from a group of humans and AI agents, not legal advice.

What applies, and when

Two duties apply today. One more is coming.

Since February 2025 In force

AI literacy

Everyone who works with AI needs enough understanding to use it responsibly. Authorities can check it since August 2026.

Since August 2026 In force

Transparency

Assistants must say they are AI. AI-generated text, images, audio and video must be marked.

From 2 December 2027 Prepare

High risk

Strict duties for AI that shapes decisions about people: hiring, credit, education and essential services.

The high-risk date moved from August 2026 with the Digital Omnibus on AI. Deferred, not removed.

Where your AI sits

The Act looks at what AI is used for, not how clever it is

Prohibited

Social scoring, manipulation, emotion recognition at work. Nothing we build.

High risk

AI that ranks, screens or decides about candidates, credit or students.

Transparency

Chatbots, assistants and AI-generated content. Say it is AI, mark what it makes.

Minimal risk

Everything else, such as an agent that drafts a reply for a person to approve.

If a person reviews what the AI proposes and makes the decision, you usually stay in the minimal or transparency class. That is why every agent we build keeps a Pioneer in the loop.

Four questions

The questions your leadership team will ask

Do we know every AI system we use?

The AI Assessment produces your AI register in one to two weeks: every licence, assistant, agent and plug-in, with a risk class and a role for each. Fixed price.

Does any of it decide about people?

We screen each use case against the prohibited list and the high-risk areas. Where AI would judge candidates, credit or students, we design it as high risk from the start or keep a person in the decision.

Do people know when they talk to AI?

Every assistant we build opens with a clear notice and a path to a person. Generated media carries machine-readable marks. Logs are kept for a defined period.

Can we show our staff were trained?

Our half-day AI at work session delivers the guideline, the session and the attendance record. That is your AI literacy evidence.

Go through them for your company: book a 15-minute call

Built to leave evidence

Every step of our method leaves something you can show

The register, logs, notices and training records the Act expects come out of work you wanted anyway.

AuditAn AI register: every system, its risk class and your role.
ConnectA record of what each AI can see and do. It also serves your GDPR records.
BuildNotices, content marks, human approval and logging, built in from the start.
ActivateYour AI literacy record: guideline, session and attendance.
OptimizeA register that stays current as systems and deadlines change.

Three ways to start

Pick the step that fits where you are

AI Assessment

Fixed price, one to two weeks. The register, the risk classes, the roadmap. The first step of our method and the first step the Act asks of you.

AI at work

A half-day session. A written guideline, a session for your team and a record of who attended. The quickest way to start with us.

Keep it current

Under a fractional AI lead or managed operations, we keep the register, logs and notices up to date.

Talk it through in 15 minutes

Tell us what AI you use or plan. We'll suggest a sensible first step, with no commitment.

Where does your AI belong?

Take the free AI Readiness Check and see where your team sits today. 5 minutes, no commitment.

We do not give legal opinions and we do not certify compliance. When someone asks how you use AI, you can show them: the register, the logs, the notices and the training records. Practical guidance from a group of humans and AI agents, not legal advice. Status as of 29 October 2026.