SOSX Insurance

SOSX Insurance is coming soon. It is the SOSX engine, set up for the actuarial, risk and exposure-management teams at insurers, MGAs and Lloyd's syndicates, and it is being built now.

It will let you map your book as a risk landscape: your lines of business, the perils they are exposed to, and the providers, regions and counterparties they depend on, joined by their dependencies. On that landscape, a validated calculation will produce the loss distributions, return periods, allocation and gross, ceded and net reinsurance figures, and accumulation scenarios will show how one shared provider or peril moves the tail. The AI will propose the assumptions, with their sources, for you to approve; the calculation will produce the figures. The aim is to explain, structure and challenge, as decision support for the people who own the numbers.

• Accumulation map. It will find the providers, regions and perils that two or more of your books share, and show the 1-in-200 and the tail with and without each shared shock, ranked by how much each drives the tail, with the path drawn on the landscape.
• Loss portfolio. It will compile your books into a loss model you can read, then give you the loss distribution per line and in total, return-period tables including the 1-in-200 and the 1-in-250, allocation and a technical price.
• Reinsurance programme. It will put your current programme and its alternatives side by side, gross, ceded and net, with the cost of each set against what it does to the tail.
• Underwriting cycle. It will show how capital, price adequacy, capacity and claims inflation pull on each other under named assumptions: scenarios to explore, never a forecast.
• Emerging-risk radar. It will track emerging signals with their sources and the exposures each could reach, and turn an "if it happened" into a draft scenario for you to approve.

Underneath, four shared risk modules will serve every SOSX edition: risk quantification (each risk's frequency and severity proposed with sources, approved by a person, and the money computed), uncertainty propagation, a scenario library you approve before use, and optimisation that weighs cost against tail risk.

• Every figure will come from a validated calculation, with its inputs and sources one click away.
• When the calculation cannot support a number, you will see why instead of a number.
• Every risk landscape will be confidential from the moment it is created, and AI work on it will use private AI processing only.
• It will work with aggregates, your books and segments, never with individual policyholders.

Harbourline Specialty is invented for this example. It is not a real company, and no figures are shown.

Harbourline is a fictional UK specialty MGA. It writes cyber, marine cargo and commercial property books, with nat cat exposure on a stretch of coast. Its cyber insureds and its cargo booking platform rely on the same cloud provider; its cargo and property books both depend on one port; its property and nat cat books sit on the same coast.

SOSX Insurance will find those three shared dependencies on the landscape. The exposure manager approves a common shock for the cloud provider, and sees the 1-in-200 and the tail with and without it, so Harbourline can show its capacity providers where its accumulation sits and what drives it. It then compares the nat cat book gross and net of a per-occurrence cover. And where the calculation cannot support a figure, it says "Insufficient" and shows why.

• Actuarial, risk and exposure-management teams at insurers, MGAs and Lloyd's syndicates, first.
• Brokers and reinsurers, next.
• General insurance and specialty, at portfolio level.