TOGAF and GRC unified: one integrated management system, run agentically

How to combine enterprise architecture based on the TOGAF Standard with governance, risk and compliance into one integrated management system: one model, one cycle joining the ADM and the management system cycle, one governance, one evidence chain — and how AI agents support operations with human approval. With interactive ADM cycle, decision aid, integration navigator, integration check and CISO whitepaper.

Last updated: November 2026 · Valeri Milke, ISO 27001 & ISO 42001 Lead Auditor

You receive the download link immediately on the page and by email.

TOGAF Standard, 10th Edition × ISO management systems · one IMS
  1. AVision
  2. BBusiness
  3. CData & Apps
  4. DTechnology
  5. ESolutions
  6. FMigration
  7. GImplement
  8. HChange
1model for architecture and GRC9ADM phases in the management system0TOGAF certificates for organisations
ADM orbit of the integrated management system: the eight phases A to H revolve around the shared model. The inner ring stands for Requirements Management as the obligations register, the outer ring for the Preliminary Phase, in which the architecture capability and the management system are set up together. Illustration: VamiSec, based on the ADM cycle of the TOGAF Standard.
9ADM phases from Preliminary to H — with Requirements Management at the centre of the cycle
7Main clauses of the Harmonized Structure (4 to 10) — shared by ISO/IEC 27001, ISO 9001, ISO 22301, ISO/IEC 42001 and other standards
5 in 1Registers — asset inventory, register of information, AI inventory, records of processing activities, SBOM — as views of one model
54Mappings in the integration navigator, from ISO-HS and ISO/IEC 27001 to NIS2, DORA, the EU AI Act, CRA and GDPR

Enterprise architecture and GRC answer two sides of the same question. Architecture describes what an organisation consists of and how it changes — capabilities, processes, applications, data, technology, service providers. GRC describes what must apply, which risks are acceptable and how to evidence both. In many organisations the two disciplines work separately: with their own inventories, their own boards and their own risk language. The result is duplicated maintenance, inconsistent data and audits for which evidence has to be laboriously pieced together. With the Architecture Development Method (ADM), the Enterprise Metamodel, the Architecture Repository and architecture governance, the TOGAF® Standard from The Open Group provides the tools to change this. Combined with the Harmonized Structure of the ISO management system standards, the result is an integrated management system: inventories become views of one model, the ADM phases carry the management system cycle, and evidence is generated in the process. Because such a system contains structured data instead of mountains of documents, AI agents can work on it — with clear permissions, a preview before every change and approval by humans.

Why now: standards and obligations that demand architecture knowledge

From the modular TOGAF Standard to the full application of the Cyber Resilience Act — the dates on which architecture and GRC have to come together.

Nine key statements on integrating TOGAF and GRC

Click a card to read the key statement in detail — or expand them all at once.

Four pillars and agentic operation: what holds the integrated management system together

What connects TOGAF and GRC — for each tab, the core idea, the key work steps and the building blocks from the TOGAF Standard and the ISO standards.

One repository
  • Extend the architecture model based on the TOGAF Standard's Enterprise Metamodel with requirement, risk, control and evidence
  • Inventories and registers as filtered views instead of parallel lists
  • Define the system of record for each object type — EA tool or GRC platform
Enterprise MetamodelArchitecture RepositoryArchiMateISO/IEC 27001 A.5.9
Interactive

Which integration path is right for you?

Up to three questions establish your starting point — from an organisation with an established architecture practice but no management system to a certified ISMS that wants to bring AI agents into operations. The recommendation is a VamiSec assessment, not an audit.

Your path
  1. ?

1: Does your organisation have an established enterprise architecture practice — a maintained architecture model (in ArchiMate, for example), a repository and a board that makes architecture decisions?

1

Does your organisation have an established enterprise architecture practice — a maintained architecture model (in ArchiMate, for example), a repository and a board that makes architecture decisions?

This means more than a folder of diagrams: capabilities, applications, data and technology are modelled, owners maintain the model, and initiatives are checked against principles and the target state. Whether you follow TOGAF to the letter does not matter.

Interactive

The integrated cycle: ADM phases with GRC tasks

Select a phase in the ring or in the list to see its goal, typical pitfalls, anchors in TOGAF, standards and regulation, work steps, guiding question and evidence. At the centre — as in the ADM cycle of the TOGAF Standard — sits Requirements Management, read here as the obligations register; around the outside wraps the agentic operating layer, an addition by VamiSec.

Preliminary Phase · ISO-HS 4 and 5 · Context, leadership, principles

Preliminary: setting up the architecture capability and the management system together

In the Preliminary Phase, the TOGAF Standard establishes how an enterprise carries out architecture work: organisation, principles, tailored approach, tools. The Harmonized Structure of the ISO management system standards poses exactly these questions in clauses 4 and 5 — context, scope, leadership, policy, roles. If you settle both in a single pass, you end up with one set of principles, one board and one scope instead of two of each.

BasisTOGAF Standard, 10th Edition: ADM, Preliminary Phase; Architecture Principles; Enterprise Architecture Capability and Governance; Series Guide “Integrating Risk and Security within a TOGAF Enterprise Architecture” (G152). ISO/IEC 27001:2022, clauses 4.1–4.4 and 5.1–5.3 (Harmonized Structure).

Typical pitfalls

  • Two sets of principles, two boards, two scopes — the silos are locked in right from the start (critical)
  • Management delegates the mandate instead of issuing it — without a mandate, the board decides nothing (critical)
  • The framework is adopted unchanged instead of tailored: many artefacts, little value (medium)
  • Tool decision before metamodel — the tool dictates the terminology instead of the other way round (medium)

Anchors in TOGAF, standards and regulation

TOGAF Preliminary PhaseTOGAF Architecture PrinciplesG152 · Risk Appetite, Security PrinciplesISO-HS 4.1–4.4ISO-HS 5.1–5.3NIS2 · Accountability of management bodies

Work steps

  • Obtain a mandate from management for an integrated management system combining architecture and GRC
  • Determine context and interested parties once — including supervisory authorities, auditors and customers
  • Formulate a joint set of principles: architecture, security, data protection and compliance principles with rationale and implications
  • Define risk appetite and key risk areas — as the TOGAF Series Guide on risk and security provides for the Preliminary Phase
  • Tailor TOGAF: which phases, artefacts and level of detail are appropriate for your size
  • Set up a joint board with rules of procedure — Architecture Board and GRC steering committee in one
  • Define the core metamodel and systems of record before procuring tools
Guiding question for management and CISO

Do we have one scope, one set of principles and one board — or are we maintaining the same fundamental decisions twice?

Deliverables & evidence

  • Management mandate and rules of procedure of the joint board
  • Context analysis with interested parties and their requirements
  • Approved set of principles
  • Tailored process model
  • Scope description

Preliminary: setting up the architecture capability and the management system together

Guide

The approach in detail: from integration principle to agentic operation

Eleven chapters: starting point, TOGAF in brief, management system, the four pillars (model, cycle, governance and evidence), regulation, agentic operation, tools and roadmap — with tables, step sequences and metrics.

01Fundamentals

Two worlds, one enterprise: why architecture and GRC belong together

Enterprise Architecture and GRC talk about the same things — capabilities, processes, applications, data, technology, service providers — but ask different questions. Architecture asks how the enterprise is structured and how it should change. GRC asks what must apply in the process, which risks are tolerable and how to prove it.

In many organisations, the two disciplines have grown up separately over the years. Architecture works with an EA tool, capability maps and an Architecture Board. The CISO office and the compliance function work with a GRC tool or spreadsheets, a risk register and a steering committee for the management system. Both sides do good work — but they do it twice, and they contradict each other in the process.

Typical symptoms of separate worlds

  • The ISMS asset inventory and the architecture’s application portfolio count different numbers of systems — and nobody knows which figure is right.
  • Criticality is assessed anew each time: in risk management, in the business impact analysis and in the DORA or NIS2 project.
  • The Architecture Board and the GRC steering committee decide on the same initiatives, but one after the other and with different criteria.
  • An approved deviation from the target architecture never shows up in the risk register as a risk acceptance.
  • Before every audit, evidence is hunted down in project folders, tickets and mailboxes.
  • Obligations such as the DORA register of information or the records of processing activities, and initiatives such as an AI inventory, are set up as special projects with their own spreadsheets.
QuestionEnterprise ArchitectureGRCIntegrated
What do we have?Capability map, application and technology portfolioAsset inventory, service provider list, AI inventoryOne model, inventories as views
What must apply?Architecture principles, standardsLaws, standards, contracts, policiesOne obligations register, one set of principles
What may change?Target architecture, roadmap, Transition ArchitecturesRisk treatment plan, measuresRoadmap as the risk treatment plan
Who decides?Architecture BoardGRC steering committee, managementOne board with a mandate from management
How do we prove it?Architecture compliance reviews, Governance RepositoryAudits, evidence repositoryOne evidence chain attached to controls and elements

Then there is speed. Regulation such as NIS2, DORA, the EU AI Act and the Cyber Resilience Act demands answers that only an up-to-date architecture model can provide: Which functions depend on which service provider? Which systems process which data? Which services are affected by an incident — and all of that within hours, not weeks.

02Fundamentals

The TOGAF® Standard, 10th Edition — read through a GRC lens

The Open Group published the 10th Edition of the TOGAF Standard on 25 April 2022 and updated it with Technical Corrigendum 1 in May 2025; the revised volumes appeared in June 2025 as the “2025 Update”. The hallmark of the 10th Edition is its modular structure: a stable Fundamental Content and TOGAF Series Guides that evolve independently.

Fundamental Content documentContentRelevance for GRC
Introduction and Core ConceptsTerms, architecture services, levels of abstraction, agility, risk management as a core conceptCommon vocabulary for architecture and GRC
Architecture Development MethodPreliminary Phase, Phases A to H, Requirements ManagementCarries the management system cycle
ADM TechniquesArchitecture Principles, Stakeholder Management, Architecture Patterns, Gap Analysis, Interoperability Requirements, Business Transformation Readiness Assessment, Risk Management, Architecture Alternatives and Trade-OffsPrinciples, stakeholders and risk as shared tools
Applying the ADMIteration, architecture styles, architecture landscape, partitioningStrategic, segment and capability iterations for the annual cycle and for initiatives
Architecture ContentContent Framework and Enterprise Metamodel, artefacts, deliverables, building blocks, Enterprise Continuum, Architecture RepositoryThe shared model and its repository
Enterprise Architecture Capability and GovernanceEstablishing the architecture function, governance framework, Architecture Board, Architecture Contracts, Architecture ComplianceBoard, contracts and reviews as part of the management system

Structure based on the 10th Edition as published in the “2025 Update”. Among other topics, the Series Guides cover Business Capabilities, Value Streams, agile approaches, Architecture Roles and Skills and the integration of risk and security.

The Architecture Development Method

The ADM describes a cycle: the Preliminary Phase prepares the organisation, principles and approach; Phase A sets the frame for an iteration with the Architecture Vision and the Statement of Architecture Work; Phases B to D develop the business, data, application and technology architectures in the Architecture Definition Document; E and F consolidate gaps into work packages, Transition Architectures, the Architecture Roadmap and the Implementation and Migration Plan; G governs implementation through Architecture Contracts and architecture compliance reviews; H monitors and decides on changes. Requirements Management sits at the centre and runs through all phases. The standard explicitly regards the ADM as iterative and adaptable.

Security and risk: cross-cutting, with a guide of its own

There is no separate security phase — security is a cross-cutting topic in all architecture domains. The Risk Management chapter of the ADM Techniques describes classification, identification, assessment before and after mitigation, and monitoring in Phase G; it is aimed primarily at the risks of architecture initiatives. For enterprise and information risks, there is the Series Guide “Integrating Risk and Security within a TOGAF Enterprise Architecture” (G152), developed by The Open Group Security Forum together with The SABSA Institute. It links information security management based on ISO/IEC 27001 and enterprise risk management based on ISO 31000 to all ADM phases.

PhaseArtefacts from G152 (selection)Link to the management system
PreliminarySecurity Principles, Risk Appetite, Key Risk Areas and Business Impact Analysis, Security Resource PlanPolicy, risk criteria, resources
B BusinessSecurity Policy Architecture, Security Domain Model, Trust Framework, Risk Assessment, Business Risk Model/Risk Register, Applicable Law and Regulation Register, Applicable Control Framework RegisterRisk assessment, obligations and control registers
C Information SystemsSecurity Services Catalog, Security ClassificationClassification, controls as security services
E SolutionsRisk Mitigation PlanRisk treatment plan
G ImplementationSecurity Audit, Security Training and AwarenessAudit, competence and awareness
Requirements ManagementBusiness Attribute Profile, Control Objectives/Security Objectives, Security StandardsObjectives and requirements

Artefact names as per G152 (first edition January 2016, listed among the Series Guides of the 10th Edition). G152 refers to ISO/IEC 27001:2013 and ISO 31000:2009; the current editions date from 2022 and 2018.

For integration, it is notable how close G152 comes to the management system: a register of applicable laws and regulations and a register of applicable control frameworks belong to the business architecture there, and controls are bundled into security services and managed as building blocks. The guide itself stresses that security is not a matter of picking items from a checklist.

What TOGAF is not

  • Not a management system standard: the standard neither requires a management system nor replaces one.
  • No certificate for organisations: The Open Group certifies individuals, with credentials such as TOGAF Enterprise Architecture Foundation and Practitioner.
  • Not a catalogue of measures: which controls are needed follows from the risk assessment, standards and laws.
  • Not a tool and not a rigid waterfall model: the ADM is tailored and iterated, including in agile sprints.
  • Not free for every use: organisations use the standard under a free, non-commercial licence for their own architecture work; consulting, training and tools for third parties require a commercial licence. The licences also prohibit processing the texts with generative AI without written permission.

TOGAF® and ArchiMate® are registered trademarks of The Open Group. This page is an independent presentation by VamiSec GmbH and is not authorised by The Open Group.

03Fundamentals

The backbone: Harmonized Structure and integrated management system

An integrated management system needs a common structure. The ISO management system standards provide it with the Harmonized Structure: the same clauses, core terms and core requirements for information security, quality, business continuity, AI and data protection.

The Harmonized Structure — set out as Appendix 2 to Annex SL in the Consolidated ISO Supplement until October 2026 and part of the joint ISO/IEC Directives since then — forms the common framework of ISO/IEC 27001, ISO 9001:2026, ISO 22301, ISO/IEC 42001, ISO/IEC 27701:2025 and other standards. Clauses 4 to 10 follow the cycle of planning, implementing, checking and improving. Anyone operating several management systems maintains these clauses once and adds only the discipline-specific requirements.

ClauseContentCounterpart in the TOGAF Standard
4 ContextOrganisation and its context, interested parties, scope, management systemPreliminary Phase, Phase A, Stakeholder Management
5 LeadershipLeadership and commitment, policy, roles and authoritiesArchitecture Board, architecture principles, EA Capability and Governance
6 PlanningRisks and opportunities, objectives, planning of changesRisk Management technique, Phases A to F, Transition Architectures
7 SupportResources, competence, awareness, communication, documented informationArchitecture Roles and Skills, Architecture Repository
8 OperationOperational planning and controlPhase G with Architecture Contracts
9 Performance evaluationMonitoring and measurement, internal audit, management reviewArchitecture compliance reviews, Phase H, Architecture Board
10 ImprovementContinual improvement, nonconformity, corrective actionsPhase H, change requests

Mapping: VamiSec. The TOGAF Standard and the Harmonized Structure do not refer to each other.

What belongs in the management system besides the structure

  • ISO 31000 as the guideline for risk management — not certifiable, but the common language for risk assessment and treatment.
  • ISO 37301 for certifiable compliance management systems, where the management of obligations is to be governed beyond information security.
  • COBIT by ISACA for the governance of enterprise IT — in version 7 since November 2026; COBIT 2019 already includes a dedicated objective, “Managed Enterprise Architecture”.
  • The Three Lines Model of the Institute of Internal Auditors — introduced in 2020 and updated in July 2026 by a position paper, with the core structure unchanged — for the division of roles between the governing body, management and internal audit.
  • The GRC Capability Model by OCEG, which describes GRC as capabilities for Principled Performance®: reliably achieving objectives, addressing uncertainty and acting with integrity.

Since March 2026, ISO has had a dedicated practical guide to the integrated management system that describes how several management systems become one — in three steps: prepare, connect, integrate. For the “connect” step, TOGAF supplies the piece that the standards deliberately leave open: a model in which the management systems meet.

TOGAF thus adds to the framework of the standards a method for describing, changing and governing the enterprise as a whole. The standards require a scope, a risk assessment and an inventory — they do not say how to build the model behind them. That is precisely where the value of integration lies.

04Pillar 1

The shared metamodel: one model instead of many inventories

The heart of integration is a shared metamodel: the definition of which object types the model knows and how they relate to one another. With its Enterprise Metamodel, the TOGAF Standard provides the architecture side; with its motivation elements, ArchiMate provides the bridge to goals, principles and requirements. GRC adds risk, control and evidence.

Object typeOriginExampleBenefit for GRC
ObligationGRCMaintain the DORA register of informationSource, applicability, owners
RequirementTOGAF, ArchiMateService provider contracts with exit clausesTranslates obligations into architecture specifications
PrincipleTOGAFData has a named ownerPuts the policy into concrete terms
Business capabilityTOGAFProcess paymentsCriticality, protection requirements, critical function
Process, value streamTOGAFLoan application to disbursementRecords of processing activities, business impact analysis
Data objectTOGAFCustomer master dataClassification, data protection
Application, AI systemTOGAFCore banking system, AI assistantAsset inventory, AI inventory
Technology, locationTOGAFCloud region, data centreOutage scenarios, concentration risk
Service provider, contractTOGAF, GRCCloud provider, framework agreementRegister of information, supply chain
RiskGRCOutage of the payment service providerAssessment, treatment, acceptance
Control, building blockGRC, TOGAFRecovery tested every quarterStatement of Applicability, multiple mapping
EvidenceGRCRecord of the recovery testAudit, supervision, certification
ExceptionTOGAF, GRCTime-limited deviation with risk acceptanceDispensation and risk acceptance in one

Core metamodel based on VamiSec practice. Terms from the TOGAF Enterprise Metamodel and from ArchiMate are adopted in substance; ten to fifteen object types are usually enough to start with.

The TOGAF Series Guide on integrating risk and security recommends exactly this approach: extending existing metamodel entities with attributes for information security and risk management instead of building a second model. ArchiMate 4, current since April 2026, drops the former Constraint element — constraints are modelled as requirements.

The compliance chain in ArchiMate

ConceptArchiMate elementExample
Regulation, supervisory authorityDriver, StakeholderDORA, financial supervisor
Risk, vulnerability, gapAssessmentConcentration of critical functions on one cloud provider
Control objectiveGoal or OutcomeCompensate for a service provider outage within hours
Control measureRequirementExit plan for each critical service provider
PolicyPrincipleNo critical function without a fallback option
Implemented controlCore elements that realise the RequirementSecondary provider, tested recovery

Pattern based on The Open Group white paper W172 on modelling risk and security with ArchiMate. ArchiMate 4 no longer has a separate Constraint element — constraints are modelled as Requirements — and a gap is modelled as an Assessment.

Registers as views of the model

As soon as the objects are linked, the registers required by regulation become filtered views instead of separate lists. This does more than save maintenance effort: the views no longer contradict each other, because they are generated from the same objects.

RegisterFrameworkView of
Asset inventoryISO/IEC 27001Applications, data objects, technology, service providers with owners
Register of informationDORAService providers, contracts, services and supported functions
AI inventoryEU AI Act (classification), ISO/IEC 42001Applications with AI characteristics, purpose, classification, data sources
Records of processing activitiesGDPRProcesses, data objects, applications, recipients
SBOMCyber Resilience ActComponents and dependencies of a product per release
Overview of service providers and dependenciesNIS2, German BSI ActService providers and their services, through to the services provided
05Pillar 2

One cycle: how the ADM phases carry the management system cycle

The ADM and the cycle of the management system standards are both designed as cycles. If you interlock them, every ADM phase delivers outputs that a management system clause needs — and every iteration generates the evidence along the way.

ADM phaseHarmonized Structure clauseOutput in the integrated systemEvidence
Preliminary4, 5, 7.1–⁠7.3Mandate, set of principles, board, tailored approachRules of procedure, principles
A Vision4.2, 6.1, 6.2Objectives, stakeholders, obligations, initial risksStatement of Architecture Work
B Business6.1Capabilities with criticality, critical functionsCapability map with assessment
C Data and Applications6.1, 7.5Inventories as views, data flows, AI inventoryRegisters generated from the model
D Technology6.1Platforms, locations, service providers, dependenciesService provider register, concentration analysis
E Solutions6.1, 6.2Controls as building blocks, risk treatmentStatement of Applicability
F Migration6.2, 6.3Roadmap linked to risks and deadlinesApproved treatment plan
G Implementation8.1, 9.2Architecture Contracts, architecture compliance reviewsReview reports, list of findings
H Change9.1, 9.3, 10Metrics, management report, improvementsManagement review, corrective actions
Requirements Management4.2, 6.3Obligations register with impact analysisTraceability through to the evidence

Mapping: VamiSec. Clause numbers as per ISO/IEC 27001:2022.

The TOGAF Standard emphasises that the ADM is adapted and iterated: an enterprise can run several iterations in parallel — strategic ones for the overall picture, others for individual segments or individual capabilities. For the integrated management system, this is an advantage: a strategic iteration carries the annual cycle of objective and risk planning, while segment and capability iterations accompany initiatives and regulatory projects.

One iteration in practice

  1. TriggerNew obligation or new initiative

    Requirements Management records the trigger, maps it to the affected elements and starts an impact analysis.

  2. Phase AVision with obligations and risks

    Objectives, stakeholders, scope and initial risk assessment; the board approves the Statement of Architecture Work.

  3. Phases B–DRisk assessment on the model

    Assess the affected capabilities, applications, data, technology and service providers; record gaps as requirements.

  4. Phases E–FTreatment as a roadmap

    Select controls as building blocks, plan work packages, have the residual risks of the transition states accepted.

  5. Phase GImplement with evidence

    Architecture Contracts with security and compliance requirements, reviews with a level of conformance, evidence filed against the control.

  6. Phase HMonitor and decide

    Metrics and findings feed into the joint management report; changes start the next iteration.

06Pillar 3

One board, one set of principles, one exception process

Architecture governance and management system governance answer the same questions: who decides on the target architecture and exceptions, according to which principles, and how is compliance checked? Brought together, coordination effort falls and decisions become consistent.

The joint board

  • Mandate and rules of procedure from management — including authority for exceptions and risk acceptances up to a defined threshold.
  • Members: architecture, information security, the risk and compliance function, data protection, business units and IT operations; internal audit attends in an advisory capacity and remains independent.
  • Tasks: approve the target architecture and roadmap, adopt the Statement of Architecture Work, decide on exceptions, evaluate architecture compliance reviews, prepare the management review.
  • Rhythm: regular meetings plus a fast-track procedure for incidents and deadlines.

One set of principles

The TOGAF Standard describes principles in terms of name, statement, rationale and implications, and its example set already includes a business principle on compliance with the law. For the integrated management system, a joint set that combines architecture, security, data protection and compliance principles is worthwhile.

PrincipleStatementRationaleImplication
Compliance by DesignObligations are addressed during design, not at acceptanceCorrections after the fact cost many times as muchEvery initiative maps its obligations in Phase A
One model, one truthInventories and registers are views of one modelContradictions between lists create audit risksNo register without a link to the model
Data has ownersEvery data object has a named owner and a classificationProtection and data protection require accountabilityNo go-live without an owner
People decide on risksRisk acceptances and exceptions are made by an authorised personAccountability cannot be delegated to toolsAgents prepare, people decide

Example principles in the format of the TOGAF Standard: name, statement, rationale, implications.

One exception process

The TOGAF Standard provides for the dispensation: a time-limited authorisation by the Architecture Board to deviate from architecture specifications subject to conditions. The management system provides for risk acceptance by the risk owner. In practice, both describe the same thing. A joint exception request records the deviation, affected elements, risk assessment, compensating measures, authority and expiry date — and an expired request automatically creates a task.

07Pillar 4

Evidence in the process, risk in the architecture context

The fourth pillar connects what auditors want to see with what architecture work produces anyway: contracts, reviews, decisions and change requests. And it gives risk a fixed place — on the architecture element.

Evidence produced along the way

Output of architecture workEvidence forStorage
Statement of Architecture Work with obligations and risk objectivesObjectives and planning (ISO-HS 6.2)Governance Repository, attached to the initiative
Architecture Contract with security and compliance requirementsOperational control (8.1), security in projects (Annex A 5.8)Attached to the initiative and the controls
Compliance Assessment from a reviewMonitoring and internal audit (9.1, 9.2)Attached to the control and the element
Dispensation with risk acceptanceRisk treatment (6.1.3)Exception register
Change Request with impact analysisPlanning of changes (6.3), change management (Annex A 8.32)Attached to the element
Minutes of the joint boardManagement review (9.3), management accountability under NIS2 and DORAGovernance Repository

Clause references as per ISO/IEC 27001:2022; mapping by VamiSec.

For architecture compliance reviews, the TOGAF Standard has its own conformance terminology, used since TOGAF 9: six levels — Irrelevant, Consistent, Compliant, Conformant, Fully Conformant and Non-conformant. “Non-conformant” applies as soon as an implemented feature deviates from the specification. In the integrated management system, the level determined is stored with the control and the element, feeds into metrics and, in the event of deviations, triggers the exception or corrective action process.

Risk on the architecture element

The Risk Management technique of the TOGAF Standard describes classification, identification, assessment before and after mitigation, and monitoring — with a focus on architecture initiatives. In the integrated management system, a single methodology based on ISO 31000 or ISO/IEC 27005 applies to all risks; TOGAF provides the structure to which risks are attached and against which they are reassessed. Those who want to compare risks per capability in monetary terms add Open FAIR from The Open Group, available since May 2025 in the versions Risk Taxonomy 3.1 and Risk Analysis 2.1.

  • Phase A: identify transformation risks and mitigation measures, initial assessment of the affected capabilities.
  • Phases B to D: assess risks to capabilities, applications, data flows, technology and service providers.
  • Phase E: confirm transformation readiness and risk, define the treatment for each risk.
  • Phase F: prioritise work packages by cost, benefit and risk, have the residual risks of the transition states accepted.
  • Phases G and H: monitor risks, trigger a new iteration if the situation or the risk appetite changes.
08Application

Implementing regulation with the architecture model: NIS2, DORA, EU AI Act, CRA and GDPR

The major EU regulations demand knowledge that resides in the architecture model — and set deadlines that can only be met with an up-to-date model. The overview shows what is required and what the model contributes.

FrameworkWhat is requiredWhat the model providesStatus
NIS2 and German BSI ActRisk management measures from asset management to the supply chain, executive management’s duty to implement, oversee and attend training, notifications within 24 and 72 hoursAsset and service provider views, dependencies for impact analyses, management reportIn Germany since 6 December 2025
DORAICT risk management framework, identification of business functions, assets and dependencies, register of information, exit strategies, incident reportsCapability map with critical functions, application and technology model, register as a view, Transition ArchitecturesApplicable since 17 January 2025
EU AI ActAI literacy, transparency; for high-risk systems: risk management, data governance, logging and human oversightAI inventory with classification, data flows, human oversight as an architecture principleTransparency obligations since 2 August 2026, high-risk under Annex III from 2 December 2027
Cyber Resilience ActSecurity by design, vulnerability handling, SBOM, reporting of exploited vulnerabilities and severe incidentsProduct architecture with building blocks, SBOM per release, reporting channelsReporting obligations since 11 September 2026, fully applicable from 11 December 2027
GDPRRecords of processing activities, data protection by design, data protection impact assessmentProcesses, data objects, applications and recipients as a view; principles and building blocksApplicable since 25 May 2018

EU AI Act deadlines as amended by Regulation (EU) 2026/1744. VamiSec assessment, not legal advice.

24 hEarly warning for significant incidents under NIS2 and for actively exploited vulnerabilities under the CRA
72 hNotification of significant incidents under NIS2 and vulnerability notification under the CRA
1 modelSource for the asset inventory, register of information, AI inventory, records of processing activities and SBOM

The benefit is clearest in an emergency. Anyone who, in the event of an incident, has to submit an early warning within 24 hours needs immediate answers to three questions: Which services are affected, which customers and functions depend on them, and which service providers are involved? A well-maintained model answers them with a query; without a model, the research only begins once the clock is already ticking.

09Operation

Agentic operation: agents work on the model, people decide

An integrated management system with a structured model is the prerequisite for AI agents to contribute meaningfully. Without a model, agents remain text generators; with a model, they analyse impacts, create mappings, collect evidence and prepare decisions — under rules set by the management system itself.

Agent taskADM linkRecommended levelApproval by
Evaluate a new obligation and map it to the affected elementsRequirements ManagementProposal with approvalObligation owners
Analyse the impact of a change on risks, controls and evidencePhase HProposal with approvalElement owners, board
Map controls to multiple frameworksPhase E, Requirements ManagementExecution after approvalControl owners
Collect evidence from connected systems and map itPhase GExecution after approval, later autonomous with loggingControl owners by sampling
Report discrepancies between model and operationsPhase HAutonomous with loggingElement owners
Monitor deadlines, certificates and expiry datesPhase HAutonomous with logging—
Draft the management report and management reviewPhase HProposal with approvalCISO, board
Risk acceptance, exception, target architecture decisionPhases A, F, GNone — preparation onlyAuthorised people

VamiSec recommendation. Levels in essence: read and propose, execute after approval, autonomous with logging.

Guardrails for agents in the management system

  • A separate identity and minimal permissions for each agent — no shared accounts, no administrator rights.
  • Access only through defined tools, such as the GRC platform’s MCP server, rather than unrestricted system access.
  • A preview before every write action, approval recorded with person and timestamp.
  • A complete log of inputs, sources, proposals, approvals and actual changes.
  • Sources required: every statement backed by evidence; unsupported answers are refused rather than guessed.
  • Protection against prompt injection: supplier responses, documents and web pages are data, never instructions.
  • Security of the AI systems themselves in line with ISO/IEC 27090, published in October 2026: the guidance describes AI-specific compromises — from data poisoning to output injection — together with their detection and defence.
  • AI governance for your own agents: entry in the AI inventory, risk assessment, human oversight, measurement of error rates — ISO/IEC 42001 provides the framework.
  • Respect the licence: do not use texts of the TOGAF Standard with generative AI without written permission from The Open Group; check the licence terms of ArchiMate and other standards — agents work on your model and your data.

International guidance points in the same direction. Singapore’s Model AI Governance Framework for Agentic AI of January 2026 recommends limiting agents’ autonomy and tool access up front and securing accountability through defined approval checkpoints; the MCP specification of July 2026 allows servers to ask the user for input in the middle of a call — technically exactly the right place for a confirmation.

How VamiGRC implements the pattern

VamiGRC sees itself as an agentic integrated management system: five management systems — information security, data protection, AI, product security and business continuity — share one data set. The platform’s AI assistant answers questions with sources and refuses unsupported answers, translates questions into queries on the GRC Graph and displays the query, carries out tasks — destructive actions only after explicit confirmation — and delegates to specialised agents. The autonomy levels range from L0 (manual) through L1 (AI advises) and L2 (AI executes, a human approves) to L3 (autonomous, information only, logged) and are set per task category; a four-eyes principle applies to critical changes, and L3 steps can be paused and reversed at any time. Every input, every tool call, every decision and every write operation is logged with timestamp, model version and rationale.

Via the Claude connector, Claude also works with this data: Claude sees only what the user is allowed to see, every change first appears as a preview, nothing is written without approval, and approvals remain in VamiGRC. For the connector, this technically implements the principle of this page — AI proposes, people decide.

The direction is also visible at The Open Group: a case study from November 2025 describes how the ADM, ArchiMate and an EA tool with AI-assisted search underpin an integrated quality management system. There is no TOGAF Series Guide on AI agents yet — the guardrails therefore come from the EU AI Act, ISO/IEC 42001 and the security guidance for agentic applications.

10Implementation

Tool landscape and exchange formats

Integration is first a question of model and governance, and only then of tools. Nevertheless, interfaces and formats determine whether inventories stay in sync and agents work securely.

ComponentRole in the integrated systemStandard or format
EA toolArchitecture model, views, roadmaps, Transition ArchitecturesArchiMate; Open Group Exchange Format (v3.1, for ArchiMate 3.1/3.2)
GRC platformObligations, risks, controls, evidence, approvals, reportsOSCAL for catalogues, profiles and assessment results
CMDB and cloud interfacesReconciliation of model and operationsVendor interfaces, synchronisation
Ticketing and project systemsWork packages, changes, findings, corrective actionsTwo-way synchronisation
Agent interfaceAccess for AI assistants and agents with permissions, preview and logModel Context Protocol (MCP)
Policy as codeAutomated checking of technical specifications and configurationsFor example Open Policy Agent

A simple pattern has proven effective: the EA tool is the system of record for the architecture objects, the GRC platform for obligations, risks, controls and evidence; both use the same identifiers and synchronise via interfaces. Agents gain access through a defined interface that enforces permissions, preview and logging. Where no EA tool exists, the GRC platform initially also holds the core model.

11Implementation

Roadmap: the core in 100 days, the integrated system in twelve months

Integration is not a big bang. A staged approach has proven effective: first the mandate, the model core and one pressing register, then governance and the evidence chain, and finally agentic operation.

  1. Day 1–15Mandate and starting point

    Mandate from management, Integration check, initial scope and a core team from architecture, the CISO office, compliance and a business unit.

  2. Day 15–45Core metamodel and systems of record

    Define ten to fifteen object types, relationships, identifiers and the system of record for each object type.

  3. Day 30–75One inventory from one source

    Merge the asset inventory and the application portfolio, name owners, generate the most pressing register — for DORA or NIS2, for example — as a view.

  4. Day 45–100Obligations register and board

    Obligations register with an intake process, a joint board with rules of procedure and a joint exception process.

  5. Month 4–6Interlock the cycle

    Initiative gate with an Architecture Contract, risks on the model, the roadmap as the risk treatment plan; align reviews and internal audit.

  6. Month 6–9Evidence chain and management report

    Evidence attached to controls and elements, continuous monitoring of the first controls, a joint report to management.

  7. Month 9–12Agentic operation

    First agent tasks with autonomy levels, approvals and logging; measure effectiveness, then expand step by step.

MetricWhat it showsDirection
Duplicate rate between inventoriesWhether the model is the single sourceFalls towards zero
Share of risks linked to an elementWhether risks are anchored in the modelRises
Lead time from new obligation → scheduled measureSpeed of Requirements ManagementFalls
Share of initiatives with an architecture and compliance gateReach of governanceRises
Open and expired exceptionsDiscipline in the exception processFalls
Time for an audit sample from requirement to evidenceEnd-to-end continuity of the evidence chainFalls
Share of rejected agent proposalsQuality of agents and approvalsMonitored
Self-assessment

Integration check: how close are your architecture and GRC to being one system?

Up to 32 questions across eight areas, each referenced to the TOGAF Standard, to the Harmonized Structure of the ISO management system standards or to regulation. Your profile determines the target level and the questions; the evaluation shows maturity by area, your biggest gaps, a roadmap for 30, 90 and 180 days and the evidence you should prepare.

0 / 32 questions answered

00 / 08Your profile

Four inputs control the check. Whether you are regulated or aiming for a certificate determines the target level. Operating model, in-house development and AI use show or hide the relevant questions — the core questions on governance, the model, registers, risk, change and evidence apply to everyone.

  1. Are you subject to NIS2, DORA or KRITIS obligations — or are you aiming for a certificate such as ISO/IEC 27001, or do you already hold one?

    Regulated entities and certified organisations must demonstrate the effectiveness of their measures. “Yes” and “Not sure yet” set the target level to “Integrated & measured”; “No” sets it to “Connected & documented” and hides the question on the auditability of the evidence chain.

  2. How do you run your IT? (Select all that apply)

    Your own data centres, your own cloud environments or outsourced services and SaaS — combinations are the norm. The question on service provider dependencies appears for cloud or external sourcing, the question on exit paths only for outsourced services.

  3. Do you develop or distribute software or products with digital elements yourself?

    This also includes internal platforms that you provide to customers, and devices with firmware. “No” hides the question on Security by Design, SBOM and vulnerability handling under the Cyber Resilience Act.

  4. Do you use AI systems or AI agents — or do you want agents to support your GRC and architecture work?

    AI systems belong in the inventory and under AI governance; agents that read and write in the model need permissions, approvals and logs. “No” hides four questions on the AI inventory and agent operations; the question of which tasks agents may take on remains visible.

Target level3 — Integrated & measured

Anyone who is subject to NIS2, DORA or KRITIS obligations, or who is seeking or holds a certificate such as ISO/IEC 27001, must demonstrate the effectiveness of their measures to auditors. We therefore apply level 3 (“Integrated & measured”).

Free whitepaper · German & English

TOGAF & GRC for CISOs — combining architecture and GRC into one integrated management system and running it agentically

The guide for decision-makers: how the TOGAF Standard and the Harmonized Structure of the ISO management system standards fit together, which metamodel connects architecture and GRC objects, how the ADM phases carry the management system cycle, how the architecture model makes it easier to implement NIS2, DORA, the EU AI Act and CRA — and how AI agents support operations with approvals and a log.

Cover of the VamiSec whitepaper TOGAF & GRC for CISOs
27 pages (DE) / 27 pages (EN)PDF, free of chargeGerman & EnglishAs of 11/2026
  • Ten key statements and an executive board briefing: why separate architecture and GRC silos cost money, speed and audit certainty
  • The shared metamodel with object types and relationships — from obligation via control and building block to evidence
  • The integrated cycle: every ADM phase with GRC tasks, outputs and references to the standards, plus governance and the exception process
  • Agentic operation with autonomy levels, approvals and guardrails — plus a 100-day plan, metrics and the tool landscape
Free Download

Request Whitepaper

TOGAF & GRC for CISOs — one integrated management system, run agentically

Language version *

The CISO whitepaper: TOGAF & GRC for CISOs

Part I · Understand

Management summary and ten key statements, the TOGAF Standard from a CISO perspective — what it does and does not deliver — the Harmonized Structure as the backbone of the integrated management system, and the cost of separate architecture and GRC silos.

Part II · Integrate

The four pillars of model, cycle, governance and evidence: the shared metamodel with object types and relationships, ADM phases with GRC tasks, one board and one exception process, and risk in the architecture context with a common risk language.

Part III · Regulation

Implementing NIS2, DORA, the EU AI Act, CRA and GDPR on the basis of the architecture: which registers and evidence the model produces, how deadlines shape the roadmap and how incident impact analyses become faster.

Part IV · Run agentically

Agent roles from obligations analysis to draft reports, autonomy levels and approvals, guardrails based on the EU AI Act, ISO/IEC 42001 and OWASP — plus the tool landscape, a 100-day plan and metrics for management and auditors.

Agentic operation with VamiGRC

From the shared model to agentic operation: how the integrated management system works in VamiGRC

The integration described on this page needs a platform on which architecture and GRC objects are connected and agents can work in a controlled way. VamiGRC, the GRC platform of VamiSec GmbH, is designed as an agentic integrated management system: ISMS, privacy, AI, product security and business continuity management share one data set in the GRC Graph, in which business processes, information, assets, service providers, risks, controls and evidence are linked. A business process is maintained once and read as a processing activity, AI use case, asset, product and business impact analysis. Specialised AI agents handle routine work along autonomy levels L0 to L3, which you set for each task category: at L2 no task is closed without human approval; L3 runs autonomously with notification and logging and can be paused, reviewed and reversed at any time.

L0–L3autonomy levels per task category — from manual through AI-assisted and AI-executed with approval to autonomous with logging
GRC Graphprocesses, information, assets, service providers, risks, controls and evidence in one data set for five management systems
OSCALover 50 standards, your own frameworks as OSCAL catalogues, two to six frameworks combinable in the Framework Composer
  • Requirements Management: frameworks as OSCAL catalogues that can be combined in the Framework Composer — the basis for a joint gap analysis, audits and the Statement of Applicability.
  • Phases B to D: business processes, information, assets and service providers as linked inventories instead of parallel lists; the GRC Graph answers questions in natural language and shows the query.
  • Phases G and H: evidence linked to risks and controls in the GRC Graph, tasks with approval and a log of every prompt, tool call, decision and write operation with timestamp, model version and rationale.
  • Claude connector: Claude only sees what you are allowed to see; every change appears as a preview, nothing is written without your OK, and approvals stay in VamiGRC.

vamigrc.com — agentic GRC platform of VamiSec GmbH. Platform data on T Cloud Public in German data centres (BSI C5 attestation of T Cloud Public); the agent layer is provider-agnostic and configurable per tenant. Mapping the ADM phases to these functions is VamiSec's interpretation, not a product feature; vamigrc.com does not describe a TOGAF approach.

VamiGRC

How to implement this with VamiGRC

VamiGRC runs ISMS, privacy, AI, product security and business continuity management as one integrated management system with shared data: a business process is maintained once and read as processing activity, AI use case, asset, product and business impact analysis. AI agents handle the routine work along autonomy levels L0 to L3 — with your approval.

Glossary

Glossary: TOGAF, GRC and agentic operation

30 terms from the TOGAF Standard, the ISO management system standards, regulation and agentic operation, briefly defined in our own words; the technical term in the other language is given as an alternative.

ADMArchitecture Development Method
The architecture development method of the TOGAF Standard: an iterative cycle comprising the Preliminary Phase, Phases A to H and Requirements Management at the centre. In the integrated management system, the ADM also carries the cycle of planning, implementing, checking and improving.
AI inventoryKI-Inventar
Inventory of AI systems and AI agents with purpose, classification under the EU AI Act, data sources and owners. In the integrated model, they are managed as application components with AI attributes.
ArchiMateArchiMate-Modellierungssprache
Modelling language from The Open Group for enterprise architectures, in version 4 since April 2026. Its motivation elements such as Stakeholder, Driver, Goal, Principle and Requirement are well suited to linking obligations and controls with architecture elements.
Architecture BoardArchitekturgremium
Board that makes architecture decisions, approves deviations and monitors compliance with the target state. In the integrated variant, it takes on tasks of the GRC steering committee and prepares the management review.
Architecture Building BlockABB · Architekturbaustein
Vendor-neutral description of a required capability in the TOGAF Standard; the concrete implementation is called a Solution Building Block. Controls can be managed as reusable building blocks that meet several requirements at once.
Architecture Compliance ReviewArchitektur-Compliance-Review
Review of whether an initiative complies with the target state and the agreed requirements; the TOGAF Standard classifies the result into levels of conformance. Extended with security and compliance criteria, the review provides evidence for internal audit and certification.
Architecture ContractArchitekturvertrag
Agreement between the architecture function and the implementing project on requirements, deliverables and quality. In the integrated management system, it contains the project's security, data protection and compliance requirements.
Architecture PrinciplesArchitekturprinzipien
Principles for architecture decisions, described in the TOGAF Standard with Name, Statement, Rationale and Implications. A joint set of principles unites architecture, security, data protection and compliance principles.
Architecture RepositoryArchitektur-Repository
Store for all architecture content in the TOGAF Standard — with the Architecture Landscape, Reference Library, Standards Library, Governance Repository, Architecture Requirements Repository and Solutions Landscape. In the integrated management system, it is controlled documentation and the source of all registers.
Autonomy levelsAutonomiestufen · L0–L3
Graduated scale of how much an AI agent may execute on its own — from manual work through proposals and execution with approval to autonomous, logged execution. VamiGRC uses four levels from L0 to L3, set per task category.
Capability MapFähigkeitslandkarte
Structured overview of what an organisation must be able to do — independent of organisational structure and technology. Rated by criticality, it is the common basis for risk assessment, business impact analysis and the critical or important functions under DORA.
DispensationAusnahmegenehmigung
In the TOGAF Standard, the time-limited authorisation by the Architecture Board to deviate from architecture requirements subject to conditions. In the integrated management system, every dispensation is also a documented risk acceptance with authority and expiry date.
Enterprise ContinuumUnternehmenskontinuum
Classification framework of the TOGAF Standard that arranges architecture and solution building blocks from generic to organisation-specific. It helps to reuse reference architectures and control building blocks.
Enterprise MetamodelUnternehmensmetamodell · until TOGAF 9.2 Content Metamodel
Part of the TOGAF Content Framework that defines the entities, attributes and relationships of an architecture model. For the integrated management system, it is extended with obligation, risk, control and evidence — Series Guide G152 itself recommends adding attributes for information security and risk management to the entities.
Governance Repositoryuntil TOGAF 9.2 Governance Log
Part of the Architecture Repository in which decisions, compliance assessments and approved exceptions are recorded. Linked to controls, it becomes a source of evidence for audits.
GRC GraphGRC-Graph in VamiGRC
Data model in VamiGRC that links business processes, information, assets, service providers, risks, controls and evidence as a graph. An example of the shared model on which an integrated management system and its agents work.
Harmonized StructureHS · Harmonisierte Struktur · until October 2026 Appendix 2 to Annex SL
Common clause structure of the ISO management system standards, since October 2026 part of the joint ISO/IEC Directives: context, leadership, planning, support, operation, performance evaluation and improvement. It is the backbone of every integrated management system.
Human-in-the-LoopMensch in der Entscheidungsschleife
Operating principle in which a human reviews and approves an AI system's proposals before they take effect. For GRC agents, this means: preview, approval, log.
Integrated management systemIMS · Integriertes Managementsystem
Management system that governs several standards and obligations — such as information security, data protection, AI, business continuity and quality — with shared processes, roles and evidence. TOGAF adds the architecture model as a shared data foundation.
Model Context ProtocolMCP
Open protocol through which AI assistants and agents use defined tools and data of other systems; part of the Linux Foundation's Agentic AI Foundation since December 2025. For agentic GRC, this means: agents do not access systems freely but through approved tools with permissions and logging.
Obligations registerPflichtenregister
Register of legal, regulatory, contractual and internal requirements with source, applicability, owners and affected elements. In the integrated management system, it is the concrete form of Requirements Management.
Open FAIRFactor Analysis of Information Risk
Risk standard from The Open Group with a taxonomy and analysis method for quantifying information risks in monetary terms, since May 2025 as Risk Taxonomy 3.1 and Risk Analysis 2.1. Complements the risk management technique of the TOGAF Standard when risks are to be quantified per capability.
OSCALOpen Security Controls Assessment Language
Machine-readable NIST formats for catalogues, profiles, system descriptions and assessment results, with a mapping model for mappings between catalogues since version 1.2. This makes obligations and control mappings reusable — for people, tools and agents.
Register of informationInformationsregister (DORA)
Register of all contractual arrangements of a financial entity with ICT third-party service providers, including the functions they support. In the integrated model, it is generated from service providers, contracts and functions.
Requirements ManagementAnforderungsmanagement
Central process of the ADM: requirements are captured, assessed, maintained and passed on in all phases. In the integrated model, it becomes the obligations register with impact analysis.
Statement of ApplicabilitySoA · Erklärung zur Anwendbarkeit
Document required by ISO/IEC 27001 listing the necessary controls, their justification, their implementation status and the justified exclusions. In the integrated model, it is derived from the building block decisions of Phase E.
Statement of Architecture WorkArchitekturauftrag
Agreement at the start of an ADM cycle on scope, approach, deliverables and schedule. Extended with obligations and risk objectives, it is the starting point for an initiative's GRC planning.
Three Lines ModelModell der drei Linien
Model published by the Institute of Internal Auditors in 2020 and updated by a statement of position in July 2026 (core structure unchanged), covering the roles of the governing body, management and internal audit. It helps to assign the architecture function and GRC roles clearly.
TOGAF Series GuidesTOGAF-Leitfäden
Supplementary guides to the TOGAF Standard that are developed independently of the Fundamental Content — for example on business capabilities, value streams, agile approaches or the integration of risk and security (G152).
Transition ArchitectureÜbergangsarchitektur
Intermediate state between baseline and target architecture. Important for GRC because intermediate states carry their own risks and exit strategies under DORA can be described as Transition Architectures.
FAQ

Frequently asked questions on TOGAF, GRC and the integrated management system

Short answers to typical questions from executive management, the CISO office, enterprise architecture, compliance and internal audit.

The TOGAF Standard from The Open Group is a framework for enterprise architecture. Its 10th Edition — since May 2025 with Technical Corrigendum 1 — consists of the Fundamental Content in six documents covering the Architecture Development Method (ADM), techniques, Enterprise Metamodel, repository and governance, plus the TOGAF Series Guides for deeper topics. TOGAF is neither a law nor a management system standard and does not prescribe any security controls. It describes how to plan, decide on, implement and govern architecture — and that is exactly what makes it the ideal partner for GRC.

Not for organisations. The Open Group certifies individuals, for example in the TOGAF Enterprise Architecture programme with the Foundation and Practitioner levels. A company cannot be certified to TOGAF. Anyone who needs a certificate for their management system still has the ISMS audited against ISO/IEC 27001, or other management systems against ISO standards — architecture work based on TOGAF provides the structures and evidence for this.

Organisations can use the standard for their own architecture work under a free, non-commercial licence; consulting, training or tools for third parties require a paid commercial licence from The Open Group. Important for agentic operation: the licence terms prohibit using the texts of the standard to train generative AI, or using such tools to generate content from them, without written permission. Agents therefore work on your model and your data, not on the text of the standard. Not legal advice.

Because both talk about the same objects: business capabilities, processes, applications, data, technology and service providers. Kept separate, they produce duplicate inventories, contradictory criticality ratings, two boards and evidence that has to be searched for before every audit. Combined, every risk is linked to an architecture element, every control to a building block and every piece of evidence to a control and an element — the effect of changes on risks and obligations becomes visible immediately.

Very well, because both are designed as cycles. The Preliminary Phase and Phase A correspond to context, leadership and planning; Phases B to D deliver the risk assessment on the model; E and F plan risk treatment as a roadmap; G governs implementation and generates evidence; H monitors, evaluates and improves. Requirements Management at the centre corresponds to the ongoing handling of the requirements of interested parties. The integration navigator on this page shows the mapping for all clauses of the Harmonized Structure.

Not necessarily. What matters is a shared model with one system of record per object type. Organisations with an established EA practice keep their EA tool and connect it to the GRC platform via interfaces. Organisations without an EA practice often start with a lean model directly in the GRC platform — capabilities, processes, applications, data, technology, service providers — and decide later whether an EA tool is needed. What is important is that terms and identifiers are the same in both worlds.

Both frameworks demand knowledge that sits in the architecture model. DORA requires the identification and documentation of ICT-supported business functions, information and ICT assets and their dependencies, as well as a register of information on ICT third-party service providers. NIS2 requires risk management measures ranging from asset management to supply chain security, as well as reports within 24 and 72 hours. With a maintained model, you generate registers as views instead of compiling them, and in an incident you quickly know which services are affected.

Nothing in the requirements — a lot in the way you work. Scope, asset inventory, risk assessment, Statement of Applicability, internal audit and management review are based on the shared model and the outputs of the ADM phases. For auditors, the ISMS becomes easier to follow because the chain from requirement via control and building block to evidence is continuous. Important: documentation in the repository must meet the standard's requirements for document control — approval, versioning, retention.

Management decides, and both variants work. More important than who chairs it are the mandate, composition and rules of procedure: architecture, information security, the risk and compliance functions, data protection and the business units belong on it, while internal audit remains independent. The board decides on the target state, exceptions and risk acceptances within its authority and reports to management, which bears overall responsibility under NIS2 and DORA.

AI agents take on clearly defined tasks on the shared model: evaluating new obligations and mapping them to affected elements, analysing the impact of changes on risks and controls, mapping controls to multiple frameworks, collecting evidence from connected systems, reporting discrepancies between model and operations, and drafting reports. They work through defined tools with their own minimal permissions. Write actions appear as proposals that a responsible person approves; only low-risk routines such as monitoring deadlines run autonomously — and everything is logged.

We recommend against it. Risk acceptances, exceptions and decisions on the target state are management tasks involving authority and accountability — the standards require risk owners and management to decide. Agents prepare such decisions: they compile facts, dependencies, options and sources. Only low-risk routines such as monitoring, deadlines and data reconciliation may run autonomously — logged and able to be paused at any time.

Internal GRC agents are generally not high-risk AI systems under the EU AI Act. Nevertheless, obligations apply, such as promoting AI literacy and, since 2 August 2026, transparency obligations, for example when people interact with AI or AI generates content. Regardless of this, we recommend treating the agents like any other AI system: add them to the AI inventory, assess their risks, define human oversight and logging — ISO/IEC 42001 provides the framework for this. Not legal advice.

VamiGRC is designed as an agentic integrated management system: five management systems share one data set in the GRC Graph, in which processes, information, assets, service providers, risks, controls and evidence are linked. Specialised AI agents work along autonomy levels L0 to L3, which are set per task category; at level L2 no task closes without human approval, while at L3 agents work autonomously, logged and pausable at any time. Via the Claude connector, Claude works with the data; nothing is written without your OK. vamigrc.com does not describe a TOGAF approach — mapping the ADM phases to these functions is VamiSec's interpretation and part of our consulting.

That depends on the starting point. In our experience, first results — core metamodel, joint board, an inventory from a single source and an obligations register for one scope — can be achieved in around 100 days. A fully integrated management system with an exception process, coordinated audits, a joint management report and the first agentic use cases takes closer to twelve months. Organisations that already have both make faster progress; those building from scratch should prioritise along the most pressing obligations.

With the integration check on this page to see gaps per area, and with the decision aid for your path. Then three steps: a mandate from management, a core metamodel with systems of record and a first scope with the most urgent register, for DORA or NIS2, for example. Boards and the exception process follow; agents only come in once the model is clean.

They complement each other. TOGAF describes the approach and governance of architecture work. ArchiMate, also from The Open Group and in version 4 since April 2026, is the language in which the model is drawn and exchanged — including motivation elements for goals, principles and requirements. OSCAL from NIST makes catalogues of requirements and controls machine-readable and reusable. Open FAIR, again from The Open Group, provides a method for assessing risks quantitatively. Together, they produce a model that people and agents can read alike.

Standards & Sources

The content on this page is based on the following publicly available guides and studies.

The Open Group · 2025

The TOGAF® Standard, 10th Edition (C220)

Fundamental Content in six documents plus TOGAF Series Guides; current version including Technical Corrigendum 1 of May 2025. The online version can only be read after signing in.

The Open Group · 2022

The Open Group Announces Launch of the TOGAF® Standard, 10th Edition

Published on 25 April 2022: modular structure of Fundamental Content and Series Guides, more guidance for agile and digital initiatives.

The Open Group Security Forum, The SABSA Institute · 2016

TOGAF® Series Guide: Integrating Risk and Security within a TOGAF® Enterprise Architecture (G152)

Security and risk artefacts per ADM phase, including registers of applicable laws and control frameworks; listed among the Series Guides of the 10th Edition.

The Open Group · 2026

TOGAF® Series Guides — Overview

Guides to the 10th Edition, for example on Business Capabilities, Value Streams, agile approaches and Architecture Roles and Skills; as of November 2026, no guide on AI agents.

The Open Group · 2019

How to Model Enterprise Risk Management and Security with the ArchiMate® Language (W172)

Modelling of risks, controls and security requirements with ArchiMate.

The Open Group · 2026

ArchiMate® 4 Specification (C260)

Current version of the modelling language since 27 April 2026; the motivation elements no longer include the former Constraint.

The Open Group · 2025

Open FAIR™ Risk Taxonomy (O-RT), Version 3.1, and Risk Analysis (O-RA), Version 2.1

Quantitative assessment of information risks in monetary terms; versions dated 22 May 2025.

The Open Group · 2026

TOGAF® Standard, 10th Edition — Downloads and Licences

Free non-commercial licence for organisations, commercial licence for consulting, training and tools; prohibition on processing the texts with generative AI without permission.

The Open Group · 2025

Designing the Future of Quality with the TOGAF® Standard and AI (Y25A)

Case study dated 19 November 2025: ADM, ArchiMate and AI-assisted search for an integrated quality management system.

The Open Group · 2026

TOGAF® Certification Portfolio

Certifications for individuals, such as TOGAF Enterprise Architecture Foundation and Practitioner; no certification of organisations.

ISO/IEC · 2022

ISO/IEC 27001:2022 — Information security, cybersecurity and privacy protection — Information security management systems — Requirements

Clauses 4 to 10 following the Harmonized Structure and Annex A with 93 controls; source of the clause references on this page.

ISO/IEC · 2026

Harmonized Structure with guidance for use — joint document to the ISO/IEC Directives, Part 1 (formerly Annex SL, Appendix 2)

Common clause structure of the ISO management system standards; part of the joint ISO/IEC Directives since 5 October 2026.

ISO · 2026

Integrated Management System — A practical guide (IMS-PG)

Practical guide from ISO, dated 17 March 2026, for an integrated management system based on several standards: prepare, connect, integrate.

ISO · 2018

ISO 31000:2018 — Risk management — Guidelines

Guidelines for risk management, not certifiable; a common language for risk assessment and risk treatment.

ISO/IEC · 2023

ISO/IEC 42001:2023 — Information technology — Artificial intelligence — Management system

AI management system following the Harmonized Structure; framework for governing the organisation's own AI agents.

ISO/IEC · 2026

ISO/IEC 27090:2026 — Cybersecurity — Artificial intelligence (AI) — Addressing security threats and compromises to AI systems

Guidance on the security of AI systems, published in October 2026; supplements ISO/IEC 27001 and 27002.

European Parliament and Council · 2022

Directive (EU) 2022/2555 (NIS2)

Governance, risk management measures and reporting obligations for essential and important entities.

Federal Republic of Germany · 2025

German BSI Act (BSIG) as amended by the NIS2 Implementation and Cybersecurity Strengthening Act

German transposition of NIS2, in force since 6 December 2025.

BSI · 2026

BSI Standard 300-1 — Methodik Grundschutz++ (Grundschutz++ Methodology), Version 1.0

The information domain as assets with their dependencies, from the business process down to the room; CMDBs and service maps are permitted.

European Parliament and Council · 2022

Regulation (EU) 2022/2554 (DORA)

ICT risk management, register of information, incident reporting and testing; applicable since 17 January 2025.

European Parliament and Council · 2024

Regulation (EU) 2024/1689 (EU AI Act)

AI literacy, transparency, obligations for high-risk AI systems including human oversight and logging.

European Parliament and Council · 2026

Regulation (EU) 2026/1744 (Digital Omnibus on the EU AI Act)

Postpones the high-risk obligations: Annex III from 2 December 2027, Annex I from 2 August 2028.

European Parliament and Council · 2024

Regulation (EU) 2024/2847 (Cyber Resilience Act)

Cybersecurity requirements for products with digital elements; reporting obligations since 11 September 2026, full application from 11 December 2027.

European Parliament and Council · 2016

Regulation (EU) 2016/679 (GDPR)

Records of processing activities, data protection by design, data protection impact assessment.

NIST · 2026

NIST Open Security Controls Assessment Language (OSCAL)

Machine-readable catalogues, profiles, system descriptions and assessment results; with a dedicated mapping model since version 1.2.

Agentic AI Foundation (Linux Foundation) · 2026

Model Context Protocol — Specification 2026-07-28

Open protocol for tool access by AI agents; since July 2026 it supports follow-up questions to the user mid-call, for example for confirmations.

OWASP GenAI Security Project · 2025

OWASP Top 10 for Agentic Applications for 2026

Published on 9 December 2025: typical risks of agentic applications, such as goal hijacking, tool misuse, and identity and privilege abuse.

IMDA Singapore · 2026

Model AI Governance Framework for Agentic AI

Limit agents' autonomy and tool access in advance; secure accountability through defined approval points.

The Institute of Internal Auditors · 2026

The IIA's Three Lines Model — Statement of Position

Position paper of 8 July 2026 that updates the 2020 model; the core structure of the three lines is retained.

Software & Systems Modeling 18(3) · 2019

Mayer, N. et al.: An integrated conceptual model for information system security risk management supported by enterprise architecture management

Academic model for integrating architecture management and security risk management, based on TOGAF and ArchiMate.

VamiSec GmbH · 2026

VamiGRC — agentic integrated management system

Autonomy levels L0 to L3, GRC Graph, Claude connector with a preview before every change; information according to vamigrc.com, as of 21 November 2026.

Combine architecture and GRC into one system?

We support you from the integration check through the metamodel, joint board and registers to agentic operation with VamiGRC — pragmatic, designed for auditability and without promises that no standard can back up.