Qu'est-ce qu'un AI Worker ?

Un AI Worker est un agent IA autonome qui travaille comme un véritable collaborateur : il possède une personnalité, des compétences spécialisées et une mémoire, exécute des tâches de bout en bout et s'intègre à vos outils (Slack, Teams, Gmail, Notion…).

Combien de AI Workers sont disponibles ?

15 AI Workers spécialisés sont disponibles, couvrant le marketing, le développement, le design, la data, les ressources humaines, la finance et le support client.

Puis-je essayer gratuitement ?

Oui, un essai gratuit est inclus et ne nécessite aucun engagement. Vous pouvez ajouter ou retirer des AI Workers à tout moment et résilier quand vous le souhaitez.

Mes données sont-elles en sécurité ?

Oui. Vos données sont chiffrées de bout en bout, hébergées dans des environnements isolés, et ne sont jamais utilisées pour entraîner des modèles. Workers Delos est conforme au RGPD et certifié SOC 2.

EU AI Act 2026: What Your Compliance Roadmap Must Include

Par Sarah Klein — 2026-07-03

The EU AI Act is no longer a future possibility — it's an active regulatory framework with enforcement phases rolling out through 2026 and 2027. Most organizations underestimate the scope, overestimate their readiness, and discover critical gaps only when they begin formal assessment. Here's the practical roadmap distilled from 30+ compliance engagements.

The Clock Is Ticking

The EU AI Act is no longer a future possibility—it's an active regulatory framework with enforcement phases rolling out through 2026 and 2027. Organizations deploying AI systems in the European market face obligations that range from transparency requirements to full conformity assessments, depending on risk classification.

Our analysis of 30+ organizations preparing for compliance reveals a consistent pattern: most underestimate the scope, overestimate their readiness, and discover critical gaps only when they begin formal assessment. This roadmap distills what we've learned into a practical, action-oriented guide.

!EU AI Act Enforcement Timeline

Understanding Risk Classification

The AI Act categorizes AI systems into four risk tiers. Your obligations depend entirely on which tier your systems fall into.

!AI Risk Classification Pyramid

Unacceptable Risk (Prohibited)

Systems that manipulate human behavior through subliminal techniques, exploit vulnerabilities of specific groups, or enable social scoring by public authorities. These are banned outright. Most enterprise AI systems won't fall here—but if you're building consumer-facing applications with behavioral nudging, review carefully.

High Risk (Strict Obligations)

This is where most compliance work concentrates. High-risk systems include:

AI used in recruitment and hiring (CV screening, candidate assessment)
Credit scoring and creditworthiness assessment
Biometric identification and categorization
Critical infrastructure management
Education and vocational training assessment
Essential services access (healthcare, insurance, banking)

If your organization uses AI in any of these areas, you must complete a conformity assessment, maintain technical documentation, implement human oversight measures, and register the system in the EU database.

Limited Risk (Transparency Obligations)

Chatbots, emotion recognition systems, and AI-generated content (deepfakes, synthetic media) must disclose to users that they are interacting with AI. No conformity assessment required, but transparency is mandatory.

Minimal Risk (No Obligations)

Spam filters, inventory management, video game AI—most routine enterprise AI falls here. No specific obligations under the Act.

Your Compliance Roadmap: 6 Steps

Step 1: Inventory Your AI Systems (Weeks 1-3)

Before you can comply, you need to know what you have. Catalog every AI system in use, including:

System name and purpose
Vendor or internal development
Data sources used
User groups affected
Risk classification (preliminary)

In our experience, most organizations discover 30-50% more AI systems than they initially estimated. Shadow AI—tools adopted by teams without IT oversight—is the biggest source of surprise.

Step 2: Classify Each System (Weeks 3-5)

For each system, determine its risk tier. This isn't a self-assessment you can rush. If a system touches hiring, credit, biometrics, or critical infrastructure, default to high-risk and prepare for full compliance. It's easier to downgrade later than to scramble for conformity after a deadline.

Step 3: Close Documentation Gaps (Weeks 5-12)

High-risk systems require extensive technical documentation:

System architecture and training data description
Risk management system documentation
Data governance documentation
Logging and record-keeping procedures
Human oversight measures description
Accuracy, robustness, and cybersecurity specifications

Most organizations have 40-60% of this documentation scattered across teams. Centralizing it takes 6-8 weeks. Creating it from scratch takes 12-16 weeks.

Step 4: Implement Human Oversight (Weeks 8-14)

The Act requires that high-risk AI systems can be effectively overseen by humans. This means:

Designated oversight personnel with authority to override AI decisions
Training programs for oversight staff
Interfaces that surface relevant information for human review
Procedures for contesting AI-driven decisions

Step 5: Register and Declare Conformity (Weeks 12-18)

High-risk systems must be registered in the EU database before deployment. For systems developed in-house, you'll need an internal conformity assessment. For third-party systems, verify that the provider has completed their obligations and that you, as deployer, meet your specific requirements.

Step 6: Establish Ongoing Monitoring (Ongoing)

Compliance isn't a one-time event. The Act requires post-market monitoring: tracking system performance, logging incidents, reporting serious events to authorities, and updating documentation when systems are modified.

Common Pitfalls We See

1. Misclassifying systems: Organizations often classify AI as "limited risk" when it touches hiring or credit—areas explicitly designated as high-risk. When in doubt, classify up.

2. Ignoring deployer obligations: Even if you buy a compliant system from a vendor, you have obligations as a deployer: human oversight, monitoring, impact assessments, and logging.

3. Underestimating shadow AI: Teams using ChatGPT, Copilot, or other tools for hiring decisions, performance reviews, or credit assessments may be unknowingly deploying high-risk AI.

4. Treating compliance as a legal exercise only: Compliance requires technical, legal, and operational alignment. A legal team alone cannot implement logging, oversight interfaces, or monitoring systems.

The Cost of Non-Compliance

Penalties under the AI Act are severe: up to €35 million or 7% of global annual turnover for prohibited practices, up to €15 million or 3% for other obligations. But the real cost is operational: systems pulled from the market, reputational damage, and the competitive disadvantage of being unable to deploy AI while competitors move forward.

Starting Now

If you haven't begun your compliance journey, start with the inventory. It's the single most valuable exercise—it reveals what you have, what you've missed, and where your gaps are. Everything else flows from that map.

The organizations that treat the AI Act as a checklist will struggle. The ones that treat it as a governance upgrade—building better documentation, oversight, and monitoring into their AI operations—will find that compliance and quality go hand in hand.