Skip to content
WhitepaperStrategy & Compliance

AI-native compliance: automating evidence for ISO 27001 and SOC 2

10 min read

Evidence collection is where compliance programs quietly die. This whitepaper outlines an engineering model — and an AI-assisted one — for making audit evidence a byproduct of normal operations.

Ask any team that has survived a first audit where the pain was, and the answer is rarely the controls themselves. It is the evidence: hunting screenshots, reconciling spreadsheets of access reviews, proving that a policy written in January was actually followed in July. Manual evidence collection turns compliance into a recurring tax on engineering — and taxes incentivize avoidance.

This whitepaper describes a model we use with clients: evidence as byproduct, pipelines as proof, and AI as the auditor's translator.

1. Evidence as byproduct

Every control should be implemented on systems that already log their own behavior. Identity providers export access reviews. CI/CD logs approval states. Vulnerability scanners archive their own runs. The control design phase should ask one question above all: when this control operates, what artifact does it produce without a human opening a tool?

  • Design the control and its evidence together — never the control alone.
  • Prefer immutable, timestamped artifacts from systems of record over human-managed files.
  • Where no system emits evidence, the missing system is itself the finding.

2. Pipelines as proof

The strongest evidence trail is a process that cannot be completed without the control. A deploy that cannot merge without a reviewed approval proves change management more convincingly than a policy that says changes must be reviewed. Shift the argument from documentation to physics: the control is not described, it is enforced — and the enforcement log is the evidence.

3. AI as the auditor's translator

This is where modern AI earns its place. A well-built assistant can cluster raw telemetry into control narratives, draft the plain-language explanation of what an artifact proves, flag evidence gaps weeks before the auditor does, and answer internal 'show me the evidence for control X' questions instantly. The work stays human-owned; the archaeology gets automated.

Two guardrails matter. First, AI-drafted narratives are verified by control owners before they enter the evidence set — the assistant drafts, engineers attest. Second, sensitive telemetry should be processed on infrastructure you control, which is exactly the kind of private AI deployment we build under our AI & SI Innovation pillar.

4. Continuous compliance as a standing capability

When evidence is automated and narratives are AI-accelerated, the annual audit collapses into a continuous review. The program's operating rhythm becomes: monitor coverage daily, remediate gaps as defects, and let the audit sample what production has been producing all along. From zero to certified — and from certified to continuously compliant.

“Auditors should sample your normal. If the audit still feels like an emergency, the system is not finished.”

Want this as a working session?

We regularly walk teams through applying this model to their own stack and framework choice — ISO/IEC 27001, SOC 2 Type I or Type II. Reach us at hello@fortuna.haus and mention 'evidence automation'.

Take this further

Want this applied to your organization?

A short readiness conversation tells you exactly where you stand and what's next.

Get whitepapers like this delivered to your inbox: