AI Governance Is an Accountability Problem
Boards are past discussing AI. Courts, regulators, and investors are asking what your oversight can prove: Does a reporting system exist? Did red flags reach the board? Does a named person stand behind each AI-assisted decision?
What Boards Need to Answer
Advisory built around a single governance question
Every engagement applies Evidence-Based Responsibility Reconstruction℠ to the same governance question — who is accountable, and can it be proven? — at a different point of exposure: the boardroom, the budget, the incident, the signature, and the workforce.
Board AI Governance Advisory
Governance built for quarterly meetings is now overseeing AI risk that moves daily. Ongoing counsel for boards closing the gap between awareness and readiness.
Learn more →CFO AI Investment Advisory
Everyone pressuring you to spend on AI has a stake in the answer. Independent counsel for the CFO deciding what to fund, defend, or kill — tied to no vendor and no adoption agenda.
Learn more →Algorithmic Accountability Risk
When an AI agent finds a route no one authorized, "the AI went rogue" is the story that lets everyone off the hook. We reconstruct who approved, who could refuse, and who priced the risk.
Learn more →Name Standard℠ Advisory
The AI did the work; a human signed it. The Name Standard℠ tests whether that sign-off was real accountability — or ceremony that won't survive scrutiny.
Learn more →AI Workforce Materiality
Stop measuring software adoption. Start governing human-capital exposure — a 90-minute paid briefing for boards, CFOs, GCs, and CHROs.
Learn more →Institutions can't govern what their systems weren't built to see
The dashboards are green. The policies are approved. And the risk is accumulating exactly where the measurement system doesn't reach.
The employees using AI in ways your policy never anticipated are the least likely to report it — so silence gets read as compliance, and the evidence base rests on voluntary disclosure by the people with the most reason to stay quiet. That is disclosure-dependent risk.
A handful of proficient employees are silently absorbing the review, correction, and judgment work AI creates — unrecognized, unbudgeted, and one resignation away from becoming a control failure. That is unmeasured human capacity.
The approval on file still describes last quarter's system. The effective system — what the AI now touches, decides, and depends on — has moved, while every report upstream stays formally accurate. That is institutional opacity.
Your controls end where your vendor's model begins — and the vendor can change that model, its data, or its behavior without triggering anything your governance process would catch. That is vendor and deployer exposure.
Hi, I'm Akilah E. Kamaria
I started my career as a data analyst, later contributing to landmark American Bar Foundation research on race and gender discrimination — published in The American Economic Review — and to foundational work on cybersecurity information sharing at the ISAO Standards Organization.
Lozen Advisory exists because institutions cannot govern what their systems were not designed to see. Menopause-at-work made that failure visible. AI workforce risk is a second observation of the same condition.
What is Evidence-Based Responsibility Reconstruction℠?
Evidence-Based Responsibility Reconstruction℠ reconstructs who held responsibility, authority, and control and what the surviving records can prove. In a Name Standard℠ engagement, six conditions test whether a human name on AI-assisted work reflects real accountability:
Time Allocation — the responsible individual has the operational hours to verify AI-assisted outputs
Review Capacity — the reviewer has the diagnostic source data to actively validate outputs
Information Access — clear line of sight into content provenance and vendor-utilized models
Documentation Infrastructure — human validation logged as evidence that survives reconstruction
Formal Right of Refusal — real standing to halt, escalate, or document non-compliant AI outputs
A board-ready answer to one question: when the AI gets it wrong, whose name is on it?
Evidence-Based Responsibility Reconstruction℠ applied to AI software
Signature observations
"A guardrail may limit what an AI system can do. The Name Standard℠ asks who is responsible when it does it."
"The boardroom problem with AI is attribution and accountability."
"Task authorization is not risk authorization. Humans define the destination, the system selects the route — and someone has to own the route."
Cited and published —
Is your board receiving enough evidence to govern AI risk?
Request an Evidence-Based Responsibility Reconstruction℠ briefing for board, CFO, legal, or risk leadership on accountability gaps, human-attribution exposure, and what the operating record can actually prove.