With the entry into force of the EU AI Act and the release of the Code of Best Practices for General AI in July 2025, companies face a decisive moment. Artificial Intelligence is no longer just a synonym for innovation — it is now also a matter of responsibility, ethics, and regulatory compliance.
In this article, we share the key practices that help organizations unlock the potential of AI in a safe, transparent, and effective way.
1. Governance and compliance from day one
Adopting AI should start with a solid governance framework. This means:
- Documenting the entire model lifecycle.
- Setting up processes for risk assessment and auditing.
- Ensuring compliance with European regulations such as the AI Act and the GDPR.
Clear governance reduces risks while increasing trust among clients and partners.
2. Setting clear goals and measurable value
Before starting any AI initiative, companies must ask: what problem are we solving?
Best practices include:
- Defining clear KPIs (e.g., reduced processing times or higher conversion rates).
- Starting with low-risk pilot projects, testing at a small scale before expanding.
- Continuously monitoring results to refine the strategy.
3. Dedicated teams and upskilling
AI adoption requires specialized skills. Companies that establish AI-focused teams or centers of excellence accelerate adoption more effectively.
Equally important is continuous training, tailored to each role — from sales teams using AI assistants to engineers building advanced models.
4. Ethics, transparency, and data security
Trust is one of the cornerstones of responsible AI. To build it:
- Protect data with encryption, anonymization, and strict access controls.
- Ensure the quality and diversity of datasets used in training.
- Choose solutions that provide explainability, making AI decisions easier to understand.
- Establish clear ethical guidelines to prevent bias and discrimination.
5. Ongoing monitoring and adaptation
AI models are never static. Their performance should be tracked using metrics such as accuracy, recall, and F1-score, with constant adjustments:
- Retraining models with updated data.
- Fine-tuning algorithms and parameters.
- Replacing models when they no longer deliver expected outcomes.
This continuous improvement ensures that AI remains effective and relevant over time.
Embracing AI best practices is more than a regulatory obligation — it is an opportunity to create trustworthy products, more efficient processes, and stronger customer relationships.
At Luza Technology, we believe that innovation must always go hand in hand with responsibility. That is why we support Portuguese companies in implementing AI solutions aligned with the highest standards of quality and European compliance.
Sources:
- The General-Purpose AI Code of Practice | Shaping Europe’s digital future
- What is Responsible AI – Azure Machine Learning | Microsoft Learn
- Evaluating AI models — Docs | IBM watsonx
by Marcelle Guedes, Senior Data Scientist at Luza