Turning responsible AI into practice: What really makes it work?

To put these principles into action, an organisation needs three practical enablers - concrete capabilities it must set up and maintain so we can work responsibly with AI in our daily work. In this post, we will focus on three essential building blocks:

• Human-machine collaboration
• Governance frameworks
• Data readiness

Each of these supports multiple Responsible AI principles. When these enables are used together, they enable AI tools to operate safely, ethically, and effectively in the world of product information.

Let’s begin with the enabler that puts people at the center: human-machine collaboration. Even as AI tools become more powerful, it’s the partnership between humans and machines that makes AI practical, safe, and effective.

This collaboration is particularly crucial for enabling two of the Principles of Responsible AI: robustness and explainability.

Robustness

AI may be excellent at surfacing patterns or auto-generating content, but it can still misinterpret context, make inaccurate suggestions, or produce overly confident outputs from limited data. Human oversight makes sure that these suggestions are reviewed and corrected before they ever reach the end user. For example, if an AI tool proposes an incorrect torque specification for a maintenance procedure, a technical communicator, who has knowledge about what might be reasonable, can catch and fix it before it becomes a safety risk.

Accountability
In technical documentation, decisions and actions are often distributed across multiple contributors. Governance makes sure that even when an AI tool suggests or automates something, a clearly identified person or team remains responsible.

For example, if an AI system can recommend steps for an installation procedure, then responsibility is shared across two key organisations: one responsible for the AI tool itself, and another for the content it delivers.

The organisation developing the AI system must make sure the system functions reliably and aligns with safety-critical use cases.

At the same time, the organisation responsible for the content must provide verified, up-to-date source material and establish quality requirements.

Together, they must make sure the AI tool supports safe and correct installations. They can do that by coordinated testing, governance, and shared accountability. Human review may be part of the process, but organisational accountability begins during planning, implementation, and quality assurance.

Even the best AI tools are only as good as the data they rely on. That’s why data readiness is such a critical enabler of Responsible AI. It makes sure that the information feeding into an AI system is clean, structured, relevant, and complete. This in turn supports both the system’s performance and the trust people place in it.

Two principles especially depend on strong data foundations: robustness and explainability.

Robustnesss
In technical documentation, a robust AI system needs to respond reliably, whether it’s retrieving a service procedure or helping identify the right spare part. If the underlying data is fragmented, outdated, or poorly structured, even the best AI logic can deliver misleading or incomplete results. For example, if spare part data lacks clear relationships to specific product models, AI might recommend the wrong component, leading to wasted time, or worse, a failed repair. High-quality data, structured with clear metadata, helps prevent these kinds of failures.