ISO 42001 Clause 10.1: Improve the AIMS Before and After Problems

What clause 10.1 requires in practice

Clause 10.1 is a single sentence in the standard, but it carries a continuing duty. The organisation has to keep making its AI management system fit better, provide enough and deliver its intended results. The same three terms appear in management review: suitability asks whether the system fits the organisation, adequacy whether it provides enough, and effectiveness whether it achieves its intended results.

The word continual signals repeated improvement over time, not a one-off project and not necessarily constant change. The clause does not require a particular improvement register, a number of projects or a named method. Improvement can follow a nonconformity, an audit finding, a monitoring trend, a management review decision or an opportunity noticed before anything fails. Evidence normally shows up in changed processes, decisions and evaluated results rather than in a dedicated record.

Identify the result that should improve

As an original starting method, describe what currently prevents a useful decision or result. Perhaps assessment begins too late, operating limits are difficult to find or review evidence does not reach management. Explain the desired change before choosing a new tool or form.

Improvement can simplify an arrangement while preserving necessary controls. An additional approval is not automatically better if it creates delay without adding useful judgement. Equally, removing a step for convenience can weaken the system if its purpose is not understood.

Why clause 10.1 matters to the business

An AI management system that never improves becomes stale as technology, suppliers, customer expectations and the organisation's own uses change. Teams then work around rules that no longer fit, which weakens control precisely where it matters. Continual improvement keeps governance proportionate: it can remove duplicated steps, bring assessment earlier in a project and make reporting routes easier to use, all of which reduce friction and cost. For the commercial side, a visible record of learning is a differentiator in tenders and customer reviews, because it shows the organisation treats AI governance as a living capability rather than a certificate on the wall.

How to implement clause 10.1

Choose improvements by the result they should produce and test them accurately. The steps below describe one practical approach that Sterling teams adopt. Small, evidenced improvements normally serve better than a large programme of change that nobody can evaluate afterwards. Review each one against its purpose.

1. Describe what currently prevents a good decision or result, such as late assessment or hard-to-find operating limits.
2. Gather opportunities from audits, measurements, reviews, employee comments and supplier changes, and rank them by value and effort.
3. Prefer simplification where it preserves the control purpose, and avoid adding approvals that nobody can explain.
4. Define in advance what evidence would show the change helped, and by when you will look.
5. Route changes to the AIMS through planned change, and check the effect on competence, reporting and comparisons.
6. Report interim evidence accurately, without presenting an unproven benefit as achieved.
7. Feed results to management review and retire changes that did not help.

Illustrative software example: better early information

Illustrative BPO example: making concerns easier to raise

A BPO provider's staff can report AI problems through a long form designed for technical incidents. Employees rarely use it for misleading summaries or unclear operating limits. Managers want better information about everyday performance.

The firm introduces a simpler route for initial concerns while preserving appropriate triage and investigation. It examines whether useful reports increase and whether staff receive responses. A higher report count may be a sign of improved visibility rather than deteriorating service, so evaluation considers the context.

Use several sources of learning

Audit findings, monitoring, assessment, management review and employee experience can all reveal opportunities. A supplier change or new use may show that existing arrangements no longer fit. The organisation should connect these sources rather than maintain separate logs that no one considers together.

Not every suggestion needs a large project. Select a proportionate response and consider how it affects other processes. A revised assessment method can change competence needs, reporting and historical comparisons. Use planned change where the AIMS is altered.

Evaluate the change against its purpose

Define what evidence would show improvement. A shorter approval cycle may be useful, but not if significant consequences receive less attention. A clearer instruction may reduce questions, but silence could also mean users cannot find the reporting route.

Keep the interpretation accurate. Some effects take time to observe, and a small sample may support only a limited conclusion. Management can use interim evidence without presenting an unproven benefit as a result already achieved.

Evidence and common mistakes