Answer yes if you have the capability to review output from automated decision-making technologies (ML, AI agents, etc.), perform testing to reproduce expected outcomes, and retain sufficient records of relevant input and output to support the identification of unexpected or inconsistent results.
Organisations are increasingly relying on automated decision-making technologies (e.g. machine learning , AI agents, etc.) for use cases where automated decisions may be used without human review. However, given the non-deterministic nature of some of these technologies (i.e. their output does not rely on a fixed algorithm and may therefore may vary for the same inputs), it is important for organisations to maintain confidence in their expected behaviour and to audit and test, as required, to ensure they perform as expected, and confirm that any variation can be logically explained or addressed through further refinement or tuning.
Your organisation should ensure that immutable logs of the input and output of these systems are retained, including the ability to understand the specific version of the capability used, to isolate variables during testing and investigation.
There are a variety of frameworks you may use to evaluate the output and effectiveness of automated decision-making technologies, which can be incorporated into existing processes for assessing capability functionality and performance.