AACF · Agentic AI Assurance & Control
Stay in control of autonomous AI
What happens when an AI agent can retrieve sensitive information, call tools, trigger workflows or act on delegated authority?
The AACF assessment helps decision-makers make agent capabilities, permissions, controls, oversight and evidence explicit—before growing exposure becomes an operational issue.
Enterprise risks
Agentic AI requires more than a security checklist
Delegated authority
Which data, tools, actions and decisions may agents perform? Who approves and remains accountable?
Trust boundaries
How are malicious instructions, compromised sources, excessive privileges and unsafe action chains constrained?
Evidence & response
Can actions be reconstructed, exceptions escalated and agents contained by controls not solely dependent on the same agent?
Scoped services
Start where your risk and evidence justify it
1. Readiness orientation
Exploratory conversation
Clarify the use case, current permissions, risk context and immediate control concerns. This first conversation clarifies scope and priorities before any evidence-based control assessment.
2. Focused AACF Assessment
Paid, defined scope
Targeted governance and control review for one bounded agent workflow, key evidence gaps and prioritised improvements.
3. Enterprise Assurance Programme
Tailored scope
Multi-agent risk and control mapping, evidence requirements, authorised testing, accountability and remediation roadmap with relevant specialist expertise.
Five readiness questions for leadership
- Which tools and data can the agent access?
- What actions require explicit human approval?
- Can the organisation reconstruct why an action occurred?
- How are untrusted instructions and delegated privileges contained?
- Who can independently detect, intervene and recover when controls fail?
Use these questions for orientation only. Do not send architecture diagrams, API keys, security configurations or sensitive incidents through the public enquiry form.
What your organisation receives
Assessment findings
- System and delegated authority boundaries
- Risk/control observations, limitations and evidence confidence
- Human approval and escalation gaps
- Prioritised proportionate remediation actions
Actionable next decisions
- Owners, responsibilities and dependency mapping
- Test and verification plan where authorised
- Evidence and reconstructability requirements
- Scope-based remediation and governance roadmap
Who this is for
CIOs, CISOs, AI platform owners, enterprise architects, AI governance and risk leaders responsible for deploying AI assistants, RAG workflows or autonomous agents in complex organisations.
Approach: define the decision and system boundary → agree evidence and permissions → examine design and controls → review observations with owners → plan remediation and verification.
See an illustrative AACF report
Review a clearly marked fictional scenario illustrating risk observations, human oversight, control improvements and evidence requirements. It does not disclose real client findings.
View the synthetic sample report Download English sample report (PDF)
Request a scoped AACF assessment
In “Assessment interest”, choose AACF Agentic AI Assurance & Control Assessment. Describe the business use case and your high-level concern only. Detailed technical material is discussed through an agreed secure channel after scoping.
Assessment data are handled according to the agreed scope and confidentiality arrangements. Privacy Policy.
English 
Dutch