Progressio.ai maps your documentation to each applicable EU AI Act requirement for High-Risk AI systems, assesses whether the evidence supports it, and shows the resulting evidence position.
Progressio.ai performs automated, evidence-based, requirement by requirement conformity assessments against Articles 9, 10, 11, 12, 13, 14, 15, 17 and 18 of the EU AI Act.
A requirement by requirement conformity assessment evaluates whether each applicable requirement of the EU AI Act can be objectively supported by evidence, not whether a policy, process, or document merely exists.
Submitted materials are mapped to applicable regulatory requirements, assessed for sufficiency and consistency, and translated into a clear view of strengths, partial coverage, and evidence gaps.
Map submitted evidence to each applicable requirement.
Evaluate each requirement using objective evidence.
Identify where evidence is incomplete, inconsistent, or missing.
Focus corrective action by regulatory significance and effort.
Establish a traceable snapshot for future reassessment.
Measure how the evidenced position changes over time.
Objective evidence sufficiently supports the applicable requirement.
Evidence exists but is incomplete, inconsistent, or insufficient.
Evidence is missing or does not support the applicable requirement.
Conformity assessments fail when fluent language substitutes for evidence. Progressio.ai evaluates every requirement against structured evidence, not a reassuring narrative.
Purpose-built text models evaluate evidence against each requirement. General-purpose language models are not used to decide conformance.
Documentation and the regulatory framework are decomposed into a navigable graph of fully covered, non-overlapping atomic requirements. Assessment runs as a deterministic evaluation against that structure.
We train the system on continuously evolving synthetic organisations that are diversified across all applicable domains. We do not train on any customer inputs, as their data is only in transit. Errors are minimised by priority. Tier 1 is false confidence: treating a requirement as evidenced when evidence is missing or insufficient. Tier 2 covers remaining assessment errors.
Where language models fit. After the assessment is complete, a general-purpose language model articulates remediation guidance in clear, human-readable form. It does not determine evidence states, map requirements, or decide conformance.
Expand any Article to see the specific assessment focus, evidence typically involved, and findings that often reveal hidden exposure.
Establishes whether a documented, continuous risk management system exists throughout the lifecycle of the High-Risk AI System.
Evaluates whether data governance processes support the quality and suitability of data used throughout the AI lifecycle.
Evaluates whether technical documentation sufficiently explains how the AI system was designed, developed, validated, and maintained.
Assesses whether appropriate logging and record keeping enable traceability throughout the AI system lifecycle.
Evaluates whether users receive the information necessary to understand intended use, limitations, assumptions, and operating conditions.
Assesses whether appropriate human oversight mechanisms enable effective intervention throughout operation of the AI system.
Evaluates whether the AI system achieves appropriate levels of accuracy, robustness, and cybersecurity throughout its intended lifecycle.
Assesses whether an appropriate quality management system governs the development, deployment, and maintenance of the AI system.
Evaluates whether required documentation and evidence are retained appropriately and remain available for regulatory review.
