Fixing delegated authority after Blueprint Two

NO UNDERWRITER would knowingly run a major line of business with management control data available only weeks in arrears and call it effective risk management. Yet that is what most of the London market does with delegated authority.

On the core book, pricing, reserving, capital and accumulation are challenged in something close to real time. Delegated business, now around 45% of what Lloyd’s writes and the fastest-growing part of the market, is the exception. It is seen weeks after the risk is bound, through bordereaux that arrive late and in formats that still must be reconciled, often by hand.

The biggest growth area on the book is measured in arrears, and the market has come to treat that as normal.

The end of Blueprint Two has now removed the last reason to leave it that way. The quandary is what the market does next. The instinct will be to fix a data problem by buying a system, and that instinct is wrong, because a faster bordereaux process is not sufficient to resolve what is effectively a control problem.

Tackling the control problem

There were two reasons the asymmetry persisted, and both have now expired. The first was technical. The data under delegated business genuinely was too slow and too costly to support a real-time view, so managing agents built their disciplines around the parts of the book where the data was already good.

The second was institutional. For five years the answer to delegated data was Blueprint Two, the market-wide programme that would standardise reporting and lift everyone at once, and there was little incentive to fund your own infrastructure while a central fix was on its way. On March 19, in its full-year results, Lloyd’s stood down that original vision; as it acknowledged, the programme had not delivered what was envisioned. The re-platforming continues, but the single market-wide fix is no longer the plan.

Less noticed is that the technical reason dissolved at the same time. Recent advances in AI-augmented matching now allow bordereaux data to be reconciled and translated in days rather than weeks, sitting over the systems a managing agent already runs, with no new platform and no data standard imposed on cover holders from above.

The central pieces of infrastructure that Lloyd’s already provides were never the obstacle. What was missing was a way to turn messy monthly submissions into the clean form those systems expect, at speed, and that now exists.

Which exposes an uncomfortable truth: the barriers were organisational and commercial all along. Cover holders would not invest without a mandate. Brokers stayed with the spreadsheets that had worked for years, and budgets went to compliance and growth. Delegated business was treated as someone else’s risk and someone else’s data, and so it never won the argument for investment.

With the central programme stood down, the first instinct will be to reach for a replacement — a multi-year platform implementation. That is solving a different problem from the one facing the market. The gap lies in two places, and neither needs a platform implementation.

Fixing the operating model

The first gap is the effectiveness of controls, which is solved by implementing a fit-for-purpose operating model. The second is data matching and remediation, which is solved by deploying new AI solutions over the bordereaux received and the binders issued. A platform addresses neither directly, nor a faster bordereaux process is not the same as control.

An agent can automate its reconciliation, clear the backlog, and still be unable to answer when its delegated authorities oversight manager asks for evidence that performance against each binder has actually been reviewed.

The firms that get this right will fix the operating model before they buy anything. They will establish what their own oversight regime requires and identifies the handful of data fields that carry real decision weight. Only then will they prove any AI tooling on their own book, before committing any infrastructure investment. The discipline is in the order: diagnosis first, technology investment last. It reduces costs, and it asks for more restraint than reaching for a system when a regulator is pressing.

Delegation without control has always been a hidden exposure. It stayed hidden because the data arrived too late to act upon. That has changed in 2026. The regulatory pressure has been building for years; what is new is that the standard excuse for inaction no longer holds. The managing agents that close the gap this year will do so cost-effectively and on their own terms.

The ones that keep waiting, or that mistake a faster process for genuine oversight, will be doing it under scrutiny, when the oversight manager asks for evidence that is not there.

Richard Spencer is managing partner of New Link Consulting