AI Governance
    AI Governance

    Shadow AI Audit & Governance Gap Assessment

    Your team is already using AI. The question is whether anyone is governing it.

    The gap, in numbers

    71% of organizations expect to deploy agentic AI within 12 months, yet only 21% have governance capabilities in place.

    Source: Gartner survey of 2,000+ CIOs and technology executives, October 2026.

    79% of organizations run AI agents in production, but only 2% use identity security tools built to govern them.

    Source: SailPoint, Horizons of Identity Security 2026.

    EU AI Act high-risk obligations apply from December 2, 2027, and customers, insurers and auditors are already asking how you govern AI.

    Source: EU Digital Omnibus agreement, May 2026.

    Score your company in two minutes

    0 / 18

    0 of 9 answered

    Score 0 if the answer is no, 1 if it is partly true, and 2 if it is fully true and you could prove it.

    01 · Ownership

    One named person owns AI decisions.

    No one owns it
    Someone cares, unofficially
    Named owner, known to the team

    02 · Policy

    Staff have seen a written AI acceptable-use policy.

    No policy exists
    Exists; few have seen it
    Distributed, current, understood

    03 · Inventory

    You could list every AI tool and agent in use today.

    No list exists
    Partial, out of date
    A living register, current

    04 · Data

    Staff know what data must never go into a public tool.

    Never discussed
    Some people know
    Documented, taught in onboarding

    05 · Vendors

    New AI tools go through an approval check before use.

    No check exists
    Sometimes, informally
    A defined check, always used

    06 · Human review

    There are rules for when a human must review AI output.

    No rules
    Unwritten expectations
    Written, followed, evidenced

    07 · Incidents

    There is a route for reporting an AI incident.

    No route
    People would tell a manager
    Named route, named responder

    08 · Training

    Staff have been trained to use AI safely.

    No training
    One session, long ago
    Current, in onboarding

    09 · Monitoring

    AI use is reviewed on a schedule and reported to leadership.

    Never reviewed
    Once, or ad hoc
    Monthly or quarterly, reported
    0–6 Exposed7–11 At risk12–15 Getting governed16–18 Governed

    Two ways in

    Shadow AI Audit · One week

    The question I answer

    What AI is your team actually using, what company data is going into it, and where are you exposed?

    What I deliver

    • A complete inventory of the AI tools in use, approved and unapproved
    • A data map showing which data goes into which tools
    • A risk-ranked register
    • Quick wins with owners and dates
    • A 45-minute executive readout

    Best for

    Leadership teams that suspect the real tool count is higher than the approved list, and want proof in a week.

    AI Readiness & Governance Assessment · Two to three weeks

    The question I answer

    How is AI governed in your company today, scored against what good looks like, and what do you fix first?

    What I deliver

    • A maturity score across nine governance domains
    • A gap report backed by evidence
    • A prioritized 30 / 90 / 180-day roadmap
    • An executive briefing your board can read
    • A scoped proposal for the next phase

    Best for

    Leadership teams ready to close the gap and deciding whether to do it in-house or with a fractional Chief AI Officer.

    Start with either. The assessment includes its own tool inventory, so the audit isn't a prerequisite. What comes next, a governance sprint or a fractional CAIO retainer, is scoped from the roadmap.

    What governance actually means

    Nine questions every leadership team should be able to answer.

    01 · Ownership

    Who owns AI decisions?

    What good looks like

    One named owner and a backup; the board hears about AI risk at least quarterly.

    02 · Policy

    Is there a written policy staff have seen?

    What good looks like

    An acceptable-use policy naming approved tools, banned data, and who to ask.

    03 · Inventory

    Can you list every AI tool and agent in use?

    What good looks like

    A living register: what it is, who uses it, what data it touches, who approved it.

    04 · Data

    Do people know what data must never go into a public tool?

    What good looks like

    Named data classes with handling rules, taught in onboarding.

    05 · Vendors

    Is there a check before a new tool is adopted?

    What good looks like

    A short approval step covering terms, data-processing agreement and data residency, before use, not after.

    06 · Human review

    When must a human check AI output?

    What good looks like

    Clear rules for review before output reaches a customer, a decision or a record.

    07 · Incidents

    What happens when AI goes wrong?

    What good looks like

    A named route for reporting a wrong output or data exposure, and someone who responds.

    08 · Training

    Have staff been trained to use AI safely?

    What good looks like

    At least one practical session, with the same material in onboarding.

    09 · Monitoring

    Who reviews AI use on a schedule?

    What good looks like

    Usage and incidents reported to leadership monthly or quarterly.

    What this is, and isn't

    Not surveillance: I map tool use, not people. Staff input is anonymous.

    Not legal advice: findings that touch regulated data are flagged for review with your counsel.

    Not a tool sale: governance is a management discipline, and I'm vendor-neutral.

    Frameworks

    The assessment draws on recognized frameworks, including the NIST AI Risk Management Framework, ISO/IEC 42001 and the EU AI Act.

    Find out where you stand in one week.

    Book a free scoping call