Measuring what matters to support responsible AI practices

Orange Business integrates responsible AI into project design from day one. We ensure that systems are ethical, transparent, and aligned with human values, reducing risk while building trust and long-term sustainability. 

Alongside in-house expertise, we draw on ISO/IEC 42001 to structure our responsible AI framework. This AI management system standard guides organizations in constructing an ordered way to address AI’s unique challenges, including risk, ethical considerations, regulatory issues and adaptability to balance innovation with governance and continuous improvement. This includes implementing, maintaining and continually improving AI policies.

We have also introduced standardized documentation templates, covering data sources and model design limitations to strengthen traceability and transparency. We have embedded governance into our project workflows, reviewing AI requirements at every stage of development. Sustainability considerations assess the environmental impact of AI.

Our objective is to give our teams clear guidance, usable tools and consistent processes to achieve responsible AI in practice. 

We are proud to be a trailblazer in ethical AI, aligning innovations with the EU AI Act and our own Data and AI Ethics Charter. With evolving regulatory frameworks, we need clear principles to translate regulations into practical actions. 

Our Charter is a central pillar to how we approach AI, ensuring responsible AI is embedded in our business strategy and is not treated as a separate initiative. By being proactive, we maintain legal alignment while strengthening customer confidence in our AI-driven solutions, instilling transparency and trust at all levels. 

The Charter is built around key commitments: human-centric AI, fairness, transparency, accountability, security, privacy, and environmental responsibility. These shape how AI initiatives are re-evaluated, developed and overseen, creating cohesion across teams and reinforcing customer trust that innovation is handled with care and accountability. 

Orange Business uses a ‘by design’ approach to AI, whereby intelligence is embedded in solutions from the start, not bolted on later. AI use cases are defined around business objectives, with data, architecture and governance structured upfront to support measurable outcomes. 

Transparency and traceability are built into workflows rather than treated as optional deliverables. This ensures AI solutions are responsible, scalable, and aligned with long-term business strategies.

By structuring governance in this way, The Charter brings together business, technical, legal, and compliance teams, making AI projects more consistent, controlled, and reliable.

Internal control is critical to make responsible AI tangible and credible. To this end, our AI systems undergo structured internal reviews - especially those classified as higher risk. Before deployment, we verify risk classification, documentation, and accountability in accordance with The Charter and the EU AI Act. This is then applied in practice. 

We use structured KPIs aligned with our responsible AI commitments to measure our performance against our overarching AI and data framework. These cover governance and ethical oversight, risk control, transparency, security, environmental efficiency, reuse practices, user satisfaction, and data quality and performance. 

These indicators are embedded into the delivery lifecycle through validation points and review checks. All projects must demonstrate compliance with governance, quality and performance to advance.

Indicators are analyzed and benchmarked against external references, including Stanford University and other international initiatives. Benchmarks include fairness and non-discrimination, data privacy and security and environmental impact.

Reporting dashboards are visible at operational and management levels, regionally and globally. Progress is tracked, and improvement areas are identified, ensuring responsible AI is consistently measured and managed across all projects. 

We continuously adapt our governance model and refine KPIs to better reflect emerging risks, new architectures and more autonomous use cases, for example. Our target is real-time visibility and stronger traceability across the AI lifecycle. 

Responsible AI relies on people and processes for ethical decision-making, proper oversight, and accountability. To address this, we invest in employee training programs, including AI literacy programs to raise awareness of the secure, responsible and ethical use of AI.

Advanced modules are provided for AI specialists and project leaders to understand the overarching data and AI framework. Feedback and experiences are shared to help embed responsible AI into our culture and support sustainable innovation practices. 

Agentic AI is the next big challenge for us, as its autonomous decision-making makes it much harder to track, audit, and ensure alignment with environmental and ethical goals. We are thus reinforcing human oversight mechanisms, clarifying responsibility chains and strengthening logging and monitoring capabilities to ensure autonomous behaviors are auditable and controllable. 

It is essential we anticipate governance challenges as AI becomes more autonomous, ensuring that innovation advances within a framework that enforces measurable trust, accountability, and transparent, responsible impact for customers.