Skip to content

AI AUDIT · GOVERNANCE DIAGNOSTIC

AI Audit: Know What You Actually Have Before You Scale It

Understand which AI tools and processes your company uses, who is responsible and where the gaps are. The review is scoped around your teams, systems and the decisions you need to make.

Signs you need an AI audit

  • ✕

    Teams using ChatGPT, Copilot, and 6 other tools with zero data governance. Customer data leaving your perimeter daily.

  • ✕

    You've spent on AI. You don't know if it's working. Nobody can answer what's our AI ROI.

  • ✕

    Legal flagged a potential EU AI Act high-risk system. Nobody in the company knows what that means or what to do.

  • ✕

    You're about to hire an AI vendor or sign a multi-year contract. You need an independent view before you commit.

What the audit covers

  1. 01

    AI inventory

    Map every AI tool, model, and workflow in use across the company. Shadow IT included.

  2. 02

    Governance gap analysis

    Assess data flows, access controls, model versioning, and incident response against EU AI Act and ISO 42001 standards.

  3. 03

    ROI measurement

    Tie each AI deployment to a business metric. Identify which tools deliver value and which are unused spend.

  4. 04

    Risk prioritisation

    Rank findings by blast radius. Deliver a remediation roadmap sorted by risk and effort — not a 100-point checklist.

  5. 05

    Executive report

    An executive summary of the current state, risk exposure, measurement baseline and recommended actions, delivered according to the agreed scope and timeline.

Proof-of-work

  • Designed the AI governance layer for a B2B SaaS from scratch: model risk registry, audit trails and access controls.

  • Compliance and AI audit for a law firm's legaltech platform. GDPR, data minimisation and responsible AI embedded.

  • AI audit of a training school's learning management stack. Shadow AI and data risk identified and remediated.

Scope and proposal

The audit is scoped around your company size, business units, tools, data sensitivity and required depth of review. The proposal defines the access needed, interviews, deliverables, timeline and commercial terms before work begins.

FAQ

How long does the audit take?
The timeline depends on the agreed scope, the systems involved and access to the relevant people and information. The proposal specifies interviews, review activities and the delivery date.
Do you need access to our systems?
Read-only access to logs, tool registries and vendor contracts. No code access required for governance audits.
Is this specific to EU AI Act compliance?
It covers EU AI Act, GDPR AI implications, and ISO 42001 alignment. The framing is business risk, not checkbox compliance.
What's the output?
A written audit report (20-40 pages), an executive slide deck (10 slides), and a prioritised remediation roadmap.
Do you help implement the remediation?
Optionally. The audit is self-contained. Clients can implement internally or engage for a follow-on sprint.
Who can audit a company's AI?
Someone who combines technical, regulatory and business judgement. I run these audits as a consultant: I review which AI tools are in use, how data flows through them, their exposure under the EU AI Act and GDPR, and what each use returns, and I deliver a report with prioritised risks and a remediation plan. Where a question needs a formal legal opinion, the report flags it for a lawyer to sign.
What kind of firm does AI audits for companies?
Technology and risk consultancies, compliance specialists and independent consultants. A good audit combines three views: technical (which systems exist and how data is used), regulatory (EU AI Act and GDPR) and business (what each use contributes). BRTHLS, Viktor Berthelius's consultancy, covers all three; anything that needs a formal legal opinion is flagged for a lawyer.

Go deeper

Know what you're running.

Tell me which decisions you need to make. We will define a review that gives your team the evidence to act.

Book an audit