Three propositions for UK motor

The market

Where UK motor stands in 2026

EY forecasts a net combined ratio of 108% for UK motor this year. The FCA has named claims handling its top insurance priority for the same year.

The regulator expects around 270,000 motorists to receive £200m of redress for write-offs settled below value. Its multi-firm review of vehicle valuation set out what good looks like:

• More than one trade guide behind every figure.
• Every deduction evidenced against that specific vehicle.
• No settlement letter worded so as to discourage a challenge.

Those are one problem seen from two ends. The account cannot be underwritten to a profit, and a total loss cannot be settled without an argument. Both come back to the same thing. The number is decided late, by a different function, from evidence nobody agreed in advance.

Who is writing

What we built in 2022

Mark Abbott spent twenty-five years in motor insurance, selling high-net-worth motor and household portfolio cover, mainly Chubb Masterpiece and Signature. He founded an insurance brokerage and sold it. Joel Smalley is an architect who has spent his career building systems in regulated markets.

In 2022, under Supermoney, we built Securus, a back-office and claims system. The deck named four problems as the reason it existed, in these words:

• “Discrepancies between accounting and settlement systems.”
• “Significant effort and cost of manual reconciliation between insurers and insurance brokers across financial accounts.”
• “Risk of human errors and fraud impacting regulatory compliance, re-insurance accounting and profitability when retrospectively reconciling book values with settlement values.”
• “Extended delay in claims settlement to the end customer, risk of losing business at renewal.”

Those were the right problems. The architecture was sound and the intelligence layer to run it did not exist. It exists now. The 2022 deck is published as it stands, at proposals.rubysdk.com/securus-2022.

Proposition one

The Guaranteed Value Floor

That sentence removes the retrospective valuation argument. The customer knows a minimum before anything happens. The insurer carries a known minimum liability across the book, priced from valuation data it already buys. It is unambiguously fair value under Consumer Duty, because it can only ever pay more than the current basis and never less.

Agreed Value on its own does not do this, and the markets that sell it broadly show why. Australia runs it as a mainstream consumer choice. The Insurance Council of Australia states plainly that an agreed value can end up lower than market value. It advises consumers to review the figure before every renewal, which tells you the design depends on the customer remembering. The failure mode is not depreciation. It is underinsurance, and a floor removes it, because the customer can never do worse than market value.

This was not buildable before. Agreeing a value per policy meant an engineer and a certificate, and only a high-net-worth premium carries that cost. Live valuation data at the point of quote has removed it. The number can now be set at inception, refreshed at renewal and compared at claim, from a feed rather than a visit.

Proposition two

Validation that is continuous, not retrospective

The facts a customer declares at inception are re-examined when a claim arrives. That happens months later, by a different function, looking at a different screen. The structure produces the same result every time. The decision is taken late, on incomplete evidence, at the worst moment for the customer.

Underwriting and claims should hold one live state. Securus was built on that principle in 2022, and it needed no AI to do it.

AI adds the part that was missing. It reads the statement of fact, the policy history, the claim and the supporting documents together, as one body of evidence. Today those are separate screens a handler reconciles by eye. It applies the rule the same way every time, so a correction propagates instead of reaching some handlers and not others. And it drafts the justification alongside the decision, so the reasoning exists at the moment the decision is made.

The floor in proposition one needs this. A value agreed at inception has to be a first-class attribute of the policy that both functions can see, not a note on a file. It has to refresh at renewal on its own. At claim it is two numbers compared, the higher one paid, and the working shown.

Proposition three

Real-time reinsurance finance reporting

Where a commission slides with the loss ratio, the accuracy and timeliness of loss ratio estimation is the income. It is not a reporting nicety. A reinsurer prices its uncertainty about a cedant's numbers into the terms it offers. Every treaty carries its own bordereaux format, settlement cycle and claims criteria, and each one multiplies both the reconciliation burden and that uncertainty.

A cedant can evidence loss development continuously, across every treaty, from one auditable source. That is a different negotiating position from a reconciled figure presented quarterly in arrears. Three things follow.

• A margin for uncertainty is a price. Reduce the uncertainty and the price should move.
• Commission settles accurately, rather than accruing conservatively and being trued up late.
• Under a pay-as-paid arrangement the recovery arrives sooner, which is cash rather than accounting.

This is the part of the argument that is hardest to see from the claims side, and it is usually worth the most. It is also the first thing Securus was built for.

Proof

Affidea Switzerland

This is running now, in a regulated setting with the same evidentiary and privacy obligations. Rubytech applies AI to billing and accounts receivable across Affidea's Swiss centres. It reads across systems that do not connect, handles insurance rejections and drafts the replies.

Patient data never leaves their premises. Identifiers are stripped before anything reaches the model, and a person decides every outcome. That last point is not optional in insurance either, and the design already answers it.

The engagement is set out in full at proposals.rubysdk.com/affidea.

Next

The step we are proposing

An hour, with nothing needed from your side beforehand. If the floor is worth testing, the step after that is pricing the tail, and that is a conversation with an actuary in the room.