By Anish Sinha

The insurance industry was placing bets on risk before most modern institutions were even a thought. Lloyd’s of London traces its roots to a coffee house on Tower Street in 1688, where merchants and ships’ captains gathered to share news and lay off the risk of voyages to the far corners of the world. The Hamburger Feuerkasse, the world’s first formal fire insurer, had already been operating for 12 years by then.

These institutions didn’t just survive the centuries, they shaped the architecture of global commerce around them. Their longevity is their greatest asset. It is also becoming their most significant liability.

In early February, Spanish digital insurer Tuio released an app that let customers obtain a quote directly through ChatGPT without leaving the conversation. The reaction from markets was swift. Shares in some of the world’s largest commercial insurance brokers – Willis, Aon, Marsh, Gallagher – fell sharply within days. Here in Australia, listed brokers including Steadfast have shed about 30% of their value over the past six months. The proximate cause: investors wondering whether AI might do what centuries of competition could not.

The market reaction was, in the short term, an overreaction. Commercial insurance is not home or motor cover. A business insuring a fleet of vehicles, a complex property portfolio or its professional indemnity exposure cannot simply punch its details into a chatbot and walk away with a binding policy.

The underwriting of business risk is bespoke by nature. Every company is different, risk is assessed case by case. The advice embedded in that process has genuine value. Brokers have not survived this long purely through inertia.

But the sell-off pointed to a structural vulnerability the industry has been papering over for years.

The commercial insurance market in Australia and New Zealand is worth about $25 billion. Almost all of it flows through brokers. Somewhere between 90% and 95% of the information exchanged between a business and its underwriter passes through a broker’s hands. And those hands, at most brokerages, are still shuffling PDFs.

The core problem is that brokers have become accidental custodians of enormous amounts of data, with almost none of it in a form that can be acted upon.

Policy information, renewal details, claim histories, underwriting submissions – it exists, but it lives in email threads and document attachments. There is no central system of record. When a business wants to switch brokers or simply get a second opinion, it has to start from scratch, filling in the same information it has provided before, answering questions about risks that have already been assessed, waiting for a process to grind through that could be dramatically compressed. The friction is a feature of the legacy model, not a bug. It keeps clients in place.

This is where AI changes the equation. Not by replacing the broker’s judgment – at least not yet – but by dismantling the data moat that legacy operators have built their competitive position on.

AI is extraordinarily good at one specific thing that happens to be the lifeblood of commercial insurance: transforming unstructured data into structured, queryable records.

An email chain about a property’s specifications, a PDF renewal schedule, a scanned claims history – AI can ingest all of it, extract the relevant information and create a record that can be interrogated, updated and acted upon. What currently takes hours of manual effort – across multiple parties sending near-identical documents with slight variations to one another – can be compressed dramatically.

The implication for the industry is not that brokers disappear overnight. It is that the advantage shifts quickly towards operators that are building with AI from the ground up rather than trying to retrofit it onto decades-old processes.

Large brokerages have not ignored this. But enterprise deployment of AI has a notoriously low success rate, and it is not hard to see why. The larger the organisation, the more interconnected its systems, the harder it is to introduce AI in a way that actually changes workflow rather than sitting alongside it.

The most likely path to transformation for the incumbents is acquisition: buying AI-native broking businesses that have built clean, modern stacks without the burden of legacy infrastructure.

Those businesses are coming. A new kind of broker is emerging that measures itself not on the size of its client book but on how quickly it can turn around a quote, how few questions it needs to ask a customer it has already served, and how seamlessly it can present a business’ complete risk profile to an underwriter. These are the metrics that matter in a world where the cost of building the underlying software is tending towards zero and the bottleneck is no longer engineering capacity – it is knowledge, relationships and the quality of the data layer underneath.

The coffee house on Tower Street was home to a genuine innovation: a place where information flowed freely and risk could be efficiently distributed among those willing to bear it. Now, the brokers that thrive will be the ones that build the digital equivalent – not just moving paper faster, but creating the kind of transparent, accessible, intelligent data infrastructure the industry has never had.

Those that treat the February sell-off as a blip rather than a signal risk discovering, too late, that the market was right.

Anish Sinha is the founder of Upcover, an AI-native commercial insurance platform