Frequently Asked Questions

Software is machinery: it stays substantially stable until someone changes it. An LLM environment behaves more like weather — its behavior emerges from the ongoing interaction of the model, the retrieval systems feeding it, enterprise data, user permissions, configurations, surrounding workflows, individual prompts, provider-side updates, and time itself. An organization can keep the same product name and license while the operating conditions underneath it change without any single approval event. Governing that as though it were a fixed executable product is the foundational mistake most enterprise AI governance still makes.

Drift is the way an AI environment changes over time even though the product name and license stay the same. We track at least eight distinct forms: content, model, retrieval, permission, context, workflow, economic, and evidence drift. Each one crosses a different institutional boundary — vendor governance, data governance, access control, legal and evidentiary readiness, management, and finance — which is exactly why no single department can own AI governance alone.

A system can log who used it, what was entered, and what response appeared — that's traceability. Accountability is different: it requires being able to reconstruct why a particular output occurred, and it requires a named person with real authority, information, time, and standing to have reviewed it. Logging preserves an event. It doesn't price it, contain it, or convert it into a decision someone can be held to.

Authorizing an AI agent to complete a task is not the same as authorizing every method it might discover to complete it, and it is not the same as pre-approving the financial exposure that method creates. With traditional software, the action path is substantially defined in advance. With agentic AI, humans define the destination — the system determines the route, and the full cost of that route may not be visible until after it has already been taken.

Each discipline sees a real piece of the problem, but none sees the whole of it. Legal tends to see liability, discrimination, and evidentiary duty. Business tends to see implementation, cost, and operating controls. Technology tends to see model performance and security. The actual governing questions — who relied on the system, who was qualified to catch a bad output, who had authority to intervene, what evidence survived, whose judgment carried the decision — sit in the space between those disciplines, not inside any one of them.

Corporate boards are the core use case, and the buyers we work with prepare board materials and own the underlying governance, risk, and compliance functions: General Counsel, Corporate Secretaries, Risk and Compliance leaders, Internal Audit, D&O liability insurers and underwriters, private equity and portfolio governance teams, and CFOs. Separately, senior executives whose names, approvals, or professional judgment attach to AI-assisted work are served through the Name Standard℠ curriculum at Executive AI Skills.

Lozen Advisory is a strategic advisory firm. We do not audit, implement, provide HR administration, deliver benefits administration, perform employee surveillance, or provide employee-data analysis. We also do not provide legal, medical, financial, or tax advice, diagnosis, treatment, or therapy.

No. Lozen Advisory's engagements are paid advisory work. The way to start is to request a briefing — a short intake, followed by a scoped briefing built around the specific governance or accountability question your board, CFO, or risk function is facing. Request a Briefing →

We review the request, and if there's a fit, follow up to scope a briefing around the accountability, governance, or exposure question you're facing. There's no obligation, and nothing is shared outside the inquiry.

We treat business inquiries with care, but information submitted through public website forms or general email should not be treated as confidential, privileged, or protected unless a separate written agreement says otherwise. Do not submit sensitive personal, medical, legal, or financial information through a public form. Privacy Policy →

Menopause-at-work and AI workforce materiality are not two separate offerings. They are two observations of one condition: institutions cannot govern what their systems were not designed to see. In disclosure-dependent systems, the population most likely to carry a risk is often the least likely to generate the records an organization uses to measure it — whether that's a senior employee absorbing the hidden verification burden of AI adoption, or an employee navigating a symptom their organization has no formal way to record. Menopause-at-work made that governance failure visible first. AI workforce risk is a second instance of the same structural blind spot. Menopause Legislation Tracker →