This project isn't really about AI. It's about what the actuarial profession is for — and the largest new class of consequential, uncertain, unevenly-distributed risk it has ever had the chance to take responsibility for. The tools you'll find on this site aren't new. The object they're pointed at is.
Actuaries exist to make uncertain future harm manageable: to measure it, price it, hold money against it, watch it over time, and put a credentialed name on the judgment. For three centuries that discipline built the trust that lets insurers, pensions, and banks operate at all. The object of the skill has changed many times — mortality, fire, storms, longevity, credit, catastrophe. The skill has not.
AI systems now make consequential decisions at massive volume — who gets care, credit, a fair price, a safe triage, a resolved claim. Their harms are exactly the kind actuaries were built for: uncertain, delayed, tail-heavy, and unevenly distributed across groups. Yet they are governed by software-engineering habits — a single accuracy score, a “95% confidence” — that were never designed to answer how badly, how much to set aside, who is accountable.
An accountant signs the books. A structural engineer signs the bridge. An actuary signs the insurer's reserves. When an AI system harms someone today, usually no one has signed anything. That accountability vacuum is a risk to the public — and an opening for a profession whose entire identity is putting its name on uncertain numbers and standing behind them.
Nothing on this site had to be invented. Frequency-severity, credibility, reserving for incurred-but-not-reported losses, VaR/TVaR, economic capital, actual-to-expected monitoring, disparate-impact testing, and the signed opinion are mature actuarial instruments — a century of theory and practice — simply pointed at a new object. The method mappings are the intellectual heart: each classic instrument answers a governance question the accuracy number cannot.
This expands the actuarial remit well beyond insurance. Every organization running automated decisions is operating an insurable book — a health system, a bank, an outsourcing firm, a marketing team, a software org, a government agency. The interactive tool runs the same lens over eight of them precisely to make the point: the machinery is domain-general. For the profession, that means new mandates, new roles, and new relevance in the rooms where AI is actually deployed — not adjacent to them.
Regulators are pulling in exactly this direction. The NAIC Model AI Bulletin asks insurers to govern AI as model risk; Colorado's SB21-169 and the federal ECOA fair-lending regime turn on whether an AI-driven decision has a “legitimate actuarial basis.” The language of the law is already actuarial — it is, in effect, inviting the profession in. Meeting that invitation with a rigorous, signable standard is the opportunity.
The next generation of actuaries won't only price policies and reserve for claims. They'll reserve against the black box and sign for the algorithms society increasingly runs on. That is the future actuary: the same trusted discipline, taking responsibility for the defining risk of this century. This site is one concrete, worked-through argument that the profession can — and should — step into that role.
Educational research artifact demonstrated on synthetic data — an argument about the profession's direction, not actuarial, legal, or investment advice.