Skip to content
All articles
ISO 42001 9 min read· by Lars Zimmermann

One management system, not three: ISO 9001, 27001 and 42001

Three standards, not three systems: what you need only once, where integrating 9001, 27001 and 42001 breaks, and what one audit programme for all looks like.

In short

You run one system with several scopes. Document control (7.5), internal audit (9.2), management review (9.3) and improvement (10) apply to all three standards together. The risk methodology, the Statement of Applicability per standard and the certificates stay separate. The structure is settled quickly. The real work lies in the three different scopes.

Auf Deutsch lesen: deutsche Fassung

Update of 10 November 2026. ISO 9001:2026, the sixth edition, has applied since September 2026. The clauses cited here carry the same numbers there, and nothing changes in what this article says.

The typical path looks like this: ISO 9001 has been in place for fifteen years because customers wanted it. ISO 27001 was added three years ago because a major customer asked for it in a supplier audit. And now ISO/IEC 42001 is on the table because the company uses AI and someone wants to know who answers for it. Three standards. And in many companies by now also three systems, three manuals, three audit programmes, three management reviews.

That is expensive, and the worst part is not the money. Three parallel systems mean three truths about the same business. After two years nobody knows which of them is the current one. I see this in audits regularly, and I had it set up for me in exactly this way before I knew better.

There is another way. One system, several scopes. What comes together, what has to stay separate and where integration actually breaks can be named with surprising clarity. I put the three standards side by side for this.

What you really need only once

Modern ISO management system standards follow a common basic structure. The kinship reaches down to the numbering and the clause titles. These clauses describe administrative machinery, and it makes sense to run that exactly once.

ClauseTitleOnce or per standard?
5.3Roles, responsibilities and authoritiesonce, with additional roles for AI
7.2Competenceonce, evidence per subject area
7.3Awarenessonce, content extended
7.4Communicationonce
7.5Documented informationonce, this is the biggest lever
9.2Internal auditonce, one programme for all three
9.3Management reviewonce, separate agenda items
10Improvementonce, one action register

Document control under 7.5 is the item that saves the most. One control procedure, one approval scheme, one storage location, one versioning logic. Anyone who keeps three manuals instead is guaranteed to have, after a year, one version that nobody has updated, and that reliably comes to light exactly when an auditor is sitting next to it.

The roles deserve a clarification. You do not need a third officer position. For the AI part you need named responsibility, and you need it where the decisions are really made: who approves an AI system, who switches it off, who answers for the results. An additional box in the organisation chart does not replace that.

You also keep competence under 7.2 in one matrix, but the evidence differs more than it appears. For the quality part, the evidence is usually a training course. For the AI part you need people who understand two things at once: the business process and the behaviour of the model within it. Someone who knows only one of the two can only pass the results on. This is the point at which a competence matrix stops being administration and starts making a real statement.

One audit programme instead of three

This is where it becomes concrete. ISO/IEC 27001:2022 even names the audit programme explicitly in a subclause of its own, 9.2.2. ISO 19011 in its 2026 edition describes the management of an audit programme in clause 5 and takes it for granted that an audit can have several audit criteria.

That is exactly the lever. You do not audit the same process three times against three standards. You audit the process once and apply several criteria. For supplier management it looks like this:

  • ISO 9001, clause 8.4: control of externally provided processes, products and services.
  • ISO/IEC 27001, Annex A, controls A.5.19 to A.5.22: information security in supplier relationships, in agreements, in the ICT supply chain and its monitoring.
  • ISO/IEC 42001, Annex A.10: third-party and customer relationships.

One audit interview, one auditor, three criteria, one report with three mappings. The head of purchasing explains his process once instead of three times. That is not only cheaper, it is also technically better, because the interactions become visible. A service provider that runs an AI model for you is a quality risk, an information security risk and an AI risk at the same time. Anyone who looks at that in three separate audits sees one third, three times.

One audit programme also means: each year the process that is worth it gets its turn. It is common to cover the scope over the three-year certification cycle. The standards themselves only require planned intervals, and the frequency depends on risk, on change and on history. A process in which an AI system has been introduced since the last audit belongs at the front. A process that has not moved in four years and has never had a finding can tolerate a longer interval. Justifying this prioritisation in writing is half the battle, and it is exactly what an external auditor asks about first.

There is a trap in independence. Internal auditors must not audit their own work, and with AI that quickly gets tight. In many mid-sized companies the same person is the quality management representative, drives digitalisation and helped build the AI use case. Anyone who audits their own work delivers a self-declaration, not evidence. In practice this is solved with a second internal person, with an auditor from another site or by buying in capacity for exactly this part of the programme.

The same principle applies to the management review, with one restriction. One meeting, one set of minutes, but separate agenda items. Each standard requires certain inputs and certain outputs, and they have to remain findable in the minutes. Minutes that stir everything into one paragraph save an hour and cost you a piece of evidence in the audit.

In concrete terms: the quality part needs customer satisfaction, process performance and supplier evaluation. The information security part needs incidents, the effectiveness of controls and the status of risk treatment. In my view the AI part should additionally carry something that the standard does not name explicitly in 9.3.2: the current inventory of the AI systems in use, together with the changes since the last review. A model change at the provider, a new version, a modified system prompt are changes to the system, even if nobody has touched a contract. Anyone who does not keep this agenda item notices changes only once they have consequences.

Where integration really breaks

Now the part that enthusiasm for integration likes to skip. Three places cannot be merged, and in all three I regularly see findings in audits.

First, the risk methodology. The understanding of risk in ISO/IEC 27001 looks inwards: what damage arises for the organisation and its information assets? ISO/IEC 42001 additionally requires a different line of sight in 6.1.2 and 6.1.4, and clause 8.4 makes it an explicit obligation: the AI system impact assessment, that is, the effects on individuals and groups who are exposed to the system. That is not a nuance, it is a different question.

A practical case from an internal audit: the risk method was cleanly documented, complete and consistently applied. It came one to one from the ISMS. Only damage to the company was assessed. The question of what a wrong judgement by the model means for the person affected did not appear anywhere in the assessment. Formally maintained, and missing the purpose in substance.

Second, the Statement of Applicability. Both standards with an Annex A require one, and each requires its own. Annex A of ISO/IEC 27001 and Annex A of ISO/IEC 42001 have different controls, different numbering and different purposes. A frequent finding: the Statement of Applicability for the ISMS exists in its third version, the one for the AI management system does not exist at all, because nobody noticed that it is a document in its own right.

Third, the link between controls and risks. A Statement of Applicability is not a catalogue for ticking off. It is meant to justify why a control is applicable, and that justification comes from the risk assessment. If both documents look good individually but not a single line refers from one to the other, the chain is broken. You do not see that in the single document, only when the two lie side by side.

These three findings have one thing in common, and that is the real reason I am writing them down here: each individual document is in order on its own. The risk method is clean. The Statement of Applicability is maintained. The impact assessment exists. The fault never lies in the document, it always lies in the relationship of the documents to one another. That is exactly why it survives internal checks so reliably. Anyone who goes through the documents one after the other finds nothing. Anyone who lays them side by side and draws a line from the control back to the risk and on to the person affected finds the gap in ten minutes.

A fourth point belongs here for the sake of completeness, even though it is an extension and not a break. ISO/IEC 42001 knows responsibilities along the AI life cycle that do not exist in this form in the other two standards: who decides on the use of a model, who monitors its behaviour in operation, who raises concerns and with whom. This can be attached to existing roles, but it cannot be dissolved into them. Anyone who answers these questions with the sentence that IT is responsible has not answered them.

The scope is the real work

All three standards require a scope to be determined in 4.3. In an integrated system these are three different scopes, and they almost never coincide.

The quality scope typically covers the whole company. The information security scope is often narrower, for example IT and one particular site. The AI scope follows an entirely different logic: it follows the AI life cycle. Who develops, who procures, who operates, who monitors, who decides on the basis of the results.

This has a very tangible consequence that surprises many. Under ISO/IEC 42006, the audit duration for an AI management system is not measured by the total workforce, but by the number of people who work in the AI life cycle, together with their role. A company with four hundred employees in which twelve people work with AI systems is not sized like a four-hundred-person company for the AI part. Anyone comparing quotations should ask about this.

The reverse also holds: anyone who draws the AI scope too narrowly in order to get off cheaply receives a certificate that says nothing about the part the customer is interested in. That is noticed in the supplier audit, at the latest with the question of whether the scope covers the service being purchased at all.

How I set it up

The order decides whether three systems become one or whether only the folder structure has been renamed.

  • First merge document control. One procedure, one location, one versioning logic. Without this step everything else remains cosmetic.
  • Then build one audit programme that audits processes and carries standards as criteria, not the other way round.
  • Then schedule one management review, with separate agenda items per standard and minutes in which each standard finds its inputs and outputs again.
  • Then one action register for all findings, with one column for the standard and one for the process.
  • Only after that the standard-specific parts: AI risk assessment, impact assessment, Statement of Applicability per standard.

And one more sober point to finish: you receive three certificates, not one. Certification is per standard, each with its own scope on paper. The audits can be combined in one visit, and that is common, but the requirements of each standard have to be demonstrated individually. Anyone who goes in expecting an integrated system to be one third of the effort will be disappointed. The realistic savings are where this text puts them: in administration, in the audit programme, in document control. That is enough to be worth it.

Share: LinkedIn E-Mail

Frequently asked questions

Do I need three internal audits for three standards?+

No. You need one audit programme. An audit can have several audit criteria, so it can cover several standards at once. You audit the process once and map the findings to the relevant clauses of each standard. That saves time and makes interactions visible.

Is one Statement of Applicability enough for ISO 27001 and ISO 42001?+

No. Each standard has its own Annex A with its own controls and its own numbering, and each requires its own Statement of Applicability. A frequent finding is that the one for the AI management system is simply missing, because nobody recognised it as a document in its own right.

Can I reuse my ISO 27001 risk method for ISO 42001?+

Only as a framework. You can take over the structure, but not the line of sight. ISO/IEC 42001 additionally requires you to consider the effects on affected individuals and groups, and clause 8.4 turns that into an explicit impact assessment. Looking purely at damage to the company falls short.

Do I get one certificate for all three standards?+

No, you receive three certificates, each with its own scope. The audits can be combined in one visit, which is common and saves travel time and preparation. The requirements of each standard are still demonstrated individually.

How many scopes does an integrated system have?+

As many as there are standards. They rarely coincide: quality usually covers the whole company, information security is often narrower, and the AI scope follows the AI life cycle instead of the organisational structure. That is exactly where the real work of integration lies.

Author & expert review: Lars Zimmermann · ISO/IEC 42001 Senior Lead Auditor & Senior Lead Implementer · ISO/IEC 27001 Lead Auditor & Lead Implementer (PECB)

Last updated: 10 November 2026. Researched and reviewed to the best of our knowledge; not a substitute for individual legal advice.

Questions about your own case?

In a free 15-minute intro call we assess where you stand on ISO 42001, ISO 27001 and the EU AI Act, honestly and without a sales pitch.

Continue reading