EU AI Act Article 13 transparency obligations: how are teams handling technical documentation for high-risk AI components?

Article 13 of the EU AI Act requires high-risk AI systems to be designed with transparency that enables deployers to interpret outputs and use them appropriately. Specifically, technical documentation must cover the system's capabilities, limitations, and intended purpose. In practice, how are teams structuring this documentation when the AI component is embedded in a larger product (e.g. an AI-powered feature in a SaaS platform)? Are you creating separate AI-system documentation, or integrating it into existing technical specs? Also: Article 13 mentions that instructions for use must include 'the level of accuracy, robustness and cybersecurity' — but the Act doesn't define acceptable thresholds. How are your teams operationalizing these metrics for auditors?