AI Governance
Where are the important governance gaps? Assess AI use within a defined scope, examine how decisions and controls operate, and prioritize the actions that need attention before moving into implementation.
The work in context
AI adoption can outpace the organization’s view of it. Teams may use approved platforms, build their own assistants or introduce AI through existing vendors, each with different evidence and decision paths. A Governance Diagnostic establishes what can be assessed within an agreed scope and where the material gaps are. We connect findings to the tools, workflows and records inspected, then identify the decisions and implementation work needed. The result prioritizes findings without claiming every AI risk has been discovered.
A sponsor may need to know whether policies cover actual practice, where approvals stall or which controls exist only on paper. An inventory can lack ownership; a policy can lack intake; restrictions can conflict with risk decisions. The diagnostic identifies what needs attention. When accepted findings require usable policies, processes or controls, Governance Structure provides the relevant implementation scope instead.
A view of the work
Working together
Define organizational units, tools, use cases, builders and evidence to inspect. Identify discovery methods and their limits. A bounded assessment should be explicit about areas it cannot see, rather than treating the absence of a record as proof of absence.
Examine policies, decision rights, inventories and relevant technical controls. Trace how a use case is approved and what operating evidence exists. Include domain and engineering expertise where a risk cannot be understood from policy documents alone.
Review representative cases and exceptions to see how the process functions. Compare intended controls with observed behavior. Record evidence quality and uncertainty so the client can distinguish a documented weakness from a hypothesis requiring further investigation.
Classify findings and recommend actions in a practical sequence. Make client decisions, dependencies and implementation options visible. Explain which gaps can be addressed through process changes, which require technical work and which need deeper assessment before action.
Outputs and decisions
Illustrative example
An organization has an AI approval form, but approved use cases lack named owners and periodic review records. A diagnostic might inspect a representative sample, identify which obligations are unclear and distinguish a process-design gap from a record-maintenance issue. Its recommendation could prioritize ownership and review pathways before investing in new tooling. This is an assessment illustration, not an audit finding or a delivered client result. Actual priorities would depend on the assessed evidence and risks.
Illustrative scenario, not a reported client result or a product capability claim.
Scope and responsibility
Assessment is distinct from authorship, implementation and recurring operation. A diagnostic does not automatically produce a full policy baseline, procure tools or enforce technical controls. It provides no comprehensive assurance that every system has been identified or that compliance is guaranteed. Agree on sampling depth and evidence access before starting. Applicable legal or regulatory determinations require the client’s appropriate advisers. A governance kickstart that creates artifacts has a different scope from a diagnostic that identifies gaps.
Practical questions
No. The diagnostic assesses the agreed scope and evidence, records limits and recommends action. Certification, legal determinations and audit attestation are distinct responsibilities and are not implied by the assessment.
Policy authorship belongs in an agreed implementation or kickstart scope. The diagnostic can identify which policies need attention and why, without silently combining assessment with the work to put them in place.
Record the blind spot and assess what can reasonably be discovered within scope. The findings should explain confidence and limits, with additional discovery proposed where it would materially change the decision.
Next step
Bring your governance question and describe the AI use you want to assess. We’ll define the assessment and evidence needed.
Discuss this work