Compliance

Phase 1, Phase 2, then keep it running

A fixed-price gap analysis, remediation through to your audit, and upkeep after it. All in a workspace you keep.

All compliance →
Security

Testing, review and leadership

Led by a published security researcher with five CVEs. A findings report, a letter of attestation for your buyers, and a retest of your fixes.

All security →
Who we help

Prove you are secure

To the people you sell to, raise from or answer to.

All industries →
Resources

Learn the space

Original research, free tools, and plain-language guides on security and compliance, from a published security researcher.

Read the blog →
Compliance

ISO 42001 Consulting for Independents: What an AI Management System Engagement Involves

An ISO 42001 consultant helps an organization build an AI management system that can be certified against ISO/IEC 42001:2023: setting the AIMS scope, running AI risk and impact assessments, selecting Annex A controls, and connecting it all to the existing security and privacy programs. For independents, the work suits ISO 27001 implementers and privacy practitioners who learn how AI systems are built.

Why ISO 42001 work is showing up

ISO/IEC 42001 was published in December 2023 as the first certifiable management system standard for artificial intelligence. Demand comes from the same places ISO 27001 demand came from years earlier: enterprise buyers asking vendors how they govern the AI in their products, procurement questionnaires with new AI sections, and companies that want a structure before regulation catches up with them.

This post is about the shape of a 42001 engagement from the consultant's side. Our post on AI governance consultant contract work in Canada covers the wider field, the Canadian rules clients ask about and how that work reaches people. If you have built an ISMS before, much of this will feel familiar, and the differences are where the real work is.

Same structure, different subject

ISO 42001 uses the harmonized structure shared by ISO management system standards, so clauses 4 to 10 line up with ISO 27001: context, leadership, planning, support, operation, performance evaluation and improvement. Internal audit, management review, corrective action, documented information and competence all work the way an ISO 27001 practitioner expects.

What changes is the subject. An ISMS protects information. An AI management system governs how the organization develops, provides or uses AI systems responsibly, which brings in questions an ISMS never asks: whether a model's outputs are fair to the people affected, whether users know they are dealing with AI, whether the training data was appropriate, who is accountable when an automated decision goes wrong.

Scoping the AIMS

Scope is harder than in ISO 27001 because AI is often spread across a company in ways nobody has written down. Start with an inventory: which AI systems the organization builds, which it integrates from third parties, and which its staff use in daily work. Then establish the organization's role for each. ISO 42001 asks the organization to determine its roles with respect to AI systems, and the obligations look different for a company that develops models, one that offers an AI-powered product built on a third-party model, and one that only uses AI tools internally.

Common scoping decisions:

  • Which products or AI systems are in scope, and whether internal use of general-purpose AI tools is included.
  • Where the boundary sits with model providers. A client building on a third-party foundation model still owns how it uses that model, but not the model's training.
  • How the AIMS relates to an existing ISMS. Many clients run them as one integrated management system with shared processes.

AI risk assessment and impact assessment

Clause 6.1 asks for an AI risk assessment and risk treatment process, much like ISO 27001. ISO 42001 also requires an AI system impact assessment: a documented look at the potential consequences of AI systems for individuals, groups and society. These are related but different. The risk assessment is about risks to the organization meeting its AI objectives. The impact assessment looks outward at the people affected by what the system does.

In practice the consultant facilitates both. Useful habits:

  • Work system by system. A generic "AI risk" register is no help to an auditor or to the engineers.
  • Cover the whole life cycle: data sourcing, training or fine-tuning, evaluation, deployment, monitoring, change and retirement.
  • Bring engineers into the room. Risks such as data drift, prompt injection, unsafe tool use by agents or unreliable evaluation only surface when people who build the system describe it.
  • Record who is affected and how. That is the core of an impact assessment, and it is where privacy practitioners tend to be strongest.

ISO/IEC 23894 gives guidance on AI risk management, and ISO/IEC 42005 gives guidance on AI system impact assessment. Neither is required for certification, but both help structure the work.

Annex A controls and the Statement of Applicability

Like ISO 27001, ISO 42001 has an Annex A of reference controls and requires a Statement of Applicability that records which controls are needed, why, whether they are implemented, and why any are excluded. Annex A has 38 controls grouped under nine control objectives, covering:

  • Policies related to AI.
  • Internal organization, including roles, responsibilities and reporting of concerns.
  • Resources for AI systems, such as data, tooling, computing and people.
  • Assessing impacts of AI systems.
  • The AI system life cycle, from requirements and design through verification, deployment, operation and monitoring.
  • Data for AI systems, including quality, provenance and preparation.
  • Information for interested parties, such as documentation for users and reporting of adverse impacts.
  • Use of AI systems, including intended use and responsible use processes.
  • Third-party and customer relationships.

Annex B gives implementation guidance for each control, and the remaining annexes cover potential AI objectives and risk sources and use across domains. A consultant who has written an ISO 27001 SoA will recognize the mechanics. The difference is that many of these controls sit with product and engineering teams rather than IT, so the evidence lives in model cards, evaluation results, data documentation and release processes. Our ISO 42001 requirements checklist shows what the client side looks like.

How it fits with ISO 27001 and privacy

Most 42001 clients already have, or are building, ISO 27001 or SOC 2. Integration is the sensible approach: one document control process, one internal audit program, one management review covering both systems, and shared supplier management. The AI-specific pieces sit on top.

Privacy overlaps heavily. AI systems often process personal information, and Canadian clients have obligations under PIPEDA, Quebec Law 25 and provincial laws such as Alberta and BC PIPA. Law 25, for example, requires organizations to inform people when a decision about them is based exclusively on automated processing. A 42001 consultant does not have to be a privacy lawyer, but the impact assessment, data controls and transparency controls should connect to the client's privacy program rather than duplicate it.

Where NIST AI RMF comes in

The NIST AI Risk Management Framework (AI RMF 1.0, January 2023) is a voluntary framework organized around four functions: Govern, Map, Measure and Manage. NIST has also published a Generative AI Profile to apply it to generative systems. It is not certifiable, and many US clients and buyers refer to it. Clients sometimes ask whether they need 42001 or the AI RMF. They answer different questions: the AI RMF describes risk management practices, while 42001 specifies a management system an accredited certification body can audit. A useful consultant can map between the two so the client does the work once. The client-side comparison is in our post on ISO 42001 versus NIST AI RMF.

Who moves into ISO 42001 work

The practitioners who settle into this work fastest tend to come from three places:

  • ISO 27001 implementers. They already know management systems, scoping, risk treatment, SoAs, internal audit and certification audits. What they need is enough understanding of how AI systems are built, evaluated and deployed to ask good questions.
  • Privacy practitioners. Impact assessment, transparency, data governance and individual rights are their home ground. What they often need is management system discipline.
  • Technical people in ML or security. They understand the systems and their failure modes, and they need to learn governance writing and audit evidence.

The IAPP's Artificial Intelligence Governance Professional (AIGP) credential is a recognized signal for this field. ISO 42001 Lead Implementer or Lead Auditor training is also available. As with any ISO work, the consultant who prepares a client does not certify it. Certification is by an accredited certification body.

Where the TrazTech Associate Network fits

We run the TrazTech Associate Network, so weigh this accordingly. The Associate Network is how TrazTech brings independent practitioners into client work, and AI governance and security, including ISO 42001, is one of its practices. Others cover compliance readiness, privacy, offensive security, cloud, security leadership and internal audit.

Associates are independent contractors, not employees, working remotely in Canada or the US, part-time or full-time, with no exclusivity. You need at least five hours a week, and if you are employed, your employer must allow side work. There is no experience minimum; newer associates are paired with a paid senior mentor.

Applying takes about five minutes. A verified credential such as AIGP, CIPP/C, ISO 27001 Lead Auditor or CISSP puts you in the fast lane and skips the simulated assessment. Otherwise you complete a simulated client engagement of about three hours of work, with 72 hours to submit, then a selection call and portal setup with the agreement signed online.

Work is offered as fixed-scope work orders. You see the scope and the fee before you accept and can decline, and fees are fixed per milestone and agreed before work starts. Payment goes out every Friday by Interac e-Transfer or against your own invoice in CAD or USD. Larger engagements run as teams with a Lead, coverage checklists keep the work consistent, and a TrazTech Principal reviews and signs off every deliverable before the client sees it. Associates get a TrazTech email address, a verified credential, paid Lead reviews, paid mentoring, referral bonuses and a path from Associate to Senior, Lead and the core team. Work follows client demand in your practice, so approval does not guarantee volume. Every practice is listed on the careers page.

Common questions about ISO 42001 consulting

Do I need a technical AI background to be an ISO 42001 consultant?

Not a research background, but you need to understand how AI systems are built, evaluated, deployed and monitored well enough to ask the right questions. ISO 27001 and privacy practitioners usually learn this on the job alongside client engineers.

How many Annex A controls does ISO 42001 have?

ISO/IEC 42001:2023 Annex A has 38 controls under nine control objectives, from AI policies and internal organization through data, the AI system life cycle, use of AI systems and third-party relationships. Annex B gives implementation guidance.

Can a company combine ISO 42001 with ISO 27001?

Yes. Both use the same harmonized clause structure, so many organizations run an integrated management system with shared document control, internal audit and management review, adding the AI-specific risk, impact and control work on top.

Is AIGP worth getting for ISO 42001 work?

It is a recognized credential for AI governance and helps show you understand the field beyond one standard. It does not replace management system experience, which is what most 42001 engagements depend on.

Can the consultant who implements ISO 42001 also certify the client?

No. Certification is performed by an accredited certification body that is independent of whoever implemented the AI management system. The consultant supports the client through the audit but does not conduct it.

Take AI governance work as an independent associate. The TrazTech Associate Network offers fixed-scope ISO 42001 and AI governance work orders, with the scope and fee shown before you accept. Applying takes about five minutes.

Apply to the networkSee the role

Do this work with TrazTech

The TrazTech Associate Network is for independent security and compliance practitioners: scoped client work, no exclusivity, feedback after every job and a published path to the core team.

Apply to the network Open roles How it works