ISO/IEC 42001 is a management-system standard. It says establish, implement, operate, monitor and improve an AI management system, and it can be certified. It deliberately does not prescribe the full catalogue of management processes. It is not a checklist. It is not a process model. It is the specification of a system. The organisation still has to build the system.
What the standard actually asks for
The standard asks for an AI management system that systematically addresses the opportunities and risks of AI across the life cycle. It focuses on features unique to AI: continuous learning, opacity, autonomy, data dependency, and the potential for emergent behaviour. It does not prescribe how to build or evaluate the technical trustworthiness of an individual model. It asks for the organisational capability to manage AI across systems and use cases.
That capability has owners. Clause 5 requires roles, responsibilities and authorities. Process owners are identified for the processes that matter: risk, data, monitoring, supplier oversight, change. An assessor looks at whether the system runs, who owns which duty, and what evidence the processes left. An assessor does not look for a completed template.
What a list leaves unbuilt
A list of 42001 clauses, marked complete, leaves the system unbuilt. It does not assign an owner to a process. It does not say what triggers a change of intended purpose. It does not produce a record an assessor can sample. It does not feed a finding back into the process that failed. A template pack mapped to the clauses is useful inside a process. It is not the management system.
Certification does not repair that. A certificate is a decision that, at a point in time, the system met the standard. If the system was never operated, there is nothing for the decision to be about. If the system is operated and the records exist, the organisation is in a position to be assessed. The software makes the work visible so the decision can be made on evidence.
How a management system meets it
The process model sits underneath the standard. Control Objectives for AI Systems (AICOB) turns the outcomes into assigned processes. The 42001 requirements are mapped onto those processes, explicitly: one process-control activity named against a named clause. The platform is the operating management system. Documented processes and criteria are mapped to the standard. Owners are assigned. Workflows run planning, operation and review. Records are what an assessor can sample.
The same system can be configured for an integrated management system and for agentic AI. The configuration changes the depth. It does not change the form. A smaller organisation can run a lighter system against its highest-impact uses. A larger one can run the same model across a portfolio. The standard allows that. A checklist does not, because a checklist has no design factors. It has lines.
For a South African organisation the same form supports King V oversight of technology, data and information, and POPIA's requirement that the compliance framework be implemented, monitored and maintained. ISO/IEC 42001 does not replace those duties. It can supply the management-system discipline that makes them operable. Where the organisation also places high-risk systems on the Union market, the same processes can support the EU AI Act's quality management system requirement. The mapping is explicit. The work is not duplicated.
None of this certifies the organisation or declares it compliant. Those decisions sit with the assessor, the Information Regulator, a notified body where the EU AI Act applies, or the governing body itself under King V. The system makes the work visible so the decision can be made on evidence.