The follow market already belongs to the algorithm

In written evidence to the Treasury Committee, Lloyd’s described Ki as its “first fully digital and algorithmically driven syndicate”, a collaboration between Brit and Google Cloud launched in 2021. Ki offers follow-only capacity alongside nominated lead syndicates (Lloyd’s evidence, AIFS0036).

Lloyd’s listed the claimed impacts: “a reduction of time from quote-to-bind (quoting a price to agreeing the contract) from days to seconds, with availability 24/7, and improved consistency and reduced bias in quotes.”

Why follow was always going first

A subscription market splits each risk between a lead, who sets terms, and followers, who decide whether to take a line on those terms. As The Insurer put it, if “a reputable underwriter has already chosen to write a risk at a given price, the decision to follow their lead is simpler (and better suited to rules-based systems or algorithms) than that of a lead underwriter” (The Insurer).

The same piece reports that other carriers now underwrite algorithmically by following Ki’s appetite rather than building their own models, and quotes InsurX’s Gilbert Harrap: “I think we will see almost all follow underwriting move into being algorithmically delivered.”

Our view: the lead has become more important, not less

The usual fear is that algorithms replace underwriters. In follow, the opposite has happened to the lead. A follow algorithm’s decision depends on who is leading. So every judgement a lead underwriter makes now moves not just their own line, but the automated capacity keyed to them.

That is a lot of weight on one human signature. And it creates a risk the market rarely discusses: if many follow algorithms trust the same handful of leads, capacity starts moving together.

Lloyd’s said so itself

The same Lloyd’s submission warns that “the concentration of insurers using the same cloud providers and ‘AI Toolboxes’ poses systemic operational resilience risks”, and that models trained on similar data “will develop similar epistemic and quantitative processing biases.” The FCA’s Mills Review makes the same point for pricing generally: synchronised decisions can “lead to coordinated withdrawal of services from particular customer segments.”

Ki’s Alan Tua told The Insurer that “safety is embedded by the constraints that the algorithm has when it is quoting digitally.” We believe him. The question for the market is not whether one syndicate’s constraints are sound. It is whether anyone is watching how all the follow algorithms behave together when a lead gets a big risk wrong.

Algorithmic follow is underrated as a success. It is also underrated as a concentration. Both are true, and Lloyd’s should publish how it measures the second.