Ki: Algorithm-driven underwriting at Lloyd's

The Future at Lloyd’s initiative has generated plenty of headlines over the past year, as the London insurance market’s blueprint for digital transformation has continued to take shape. However, the Ki syndicate, announced one year ago by Brit, offers a working example of artificial intelligence-driven underwriting that is operating already.

Ki went live at “1/1” this year, underwriting business as the first fully digital and algorithmically-driven Lloyd’s of London syndicate. Operating on a standalone basis, it was developed between Brit, Google Cloud and University College London’s Computer Science department.

Ki underwrites what is known as ‘follow-only’ business, after another Lloyd’s syndicate’s ‘lead underwriter’ has committed the first share of capacity, at which point the broker scours the market looking for following capacity to insure the risk in full.

“We’ve had a tremendous first quarter, and the support from our twelve broker partners has been tremendous,” Dan Hearsum, managing director at Ki, tells Airmic News. “We’re ahead of our plans for this stage, and on track for our business plan for the year. The plan for 2021 is for $400m of gross written premium. We’re halfway through that, having underwritten over $200m across 1,500 risks.”

One of Ki’s central aims to streamline the administrative expense ratio – a top priority for insurers and for the Lloyd’s market in particular in recent years, which has sought to shave its costs to better compete with rivals.

After reporting its first quarter of underwriting data to Lloyd’s, Hearsum says Ki is on track to deliver its efficiency target. “Ki’s business model is designed to target a lower expense ratio than the market average and, since starting to write business, we’re on track with our plan,” he says.

Simpler and faster

Another central goal behind Ki is to keep things as simple as possible for brokers seeking following capacity for placing risks in the Lloyd’s subscription market.

Traditionally to place a risk at Lloyd’s brokers have to physically queue up at a syndicate’s ‘box’, repeating that process around the famous Underwriting Room until sufficient capacity is committed to a slip by the various syndicates to underwrite the risk. In contrast, Ki has its own online platform for to get a quote from the algorithm.

“The platform is how a broker will interact with Ki and we’ve had really positive feedback on the platform, with more than 1,200 active users,” says Hearsum. “Having logged in, by entering some data they can get a quote in under a minute. Traditionally, that might have taken three days or perhaps even a couple of weeks. Brokers can access the platform 24 hours a day, seven days a week, via their laptop, tablet or smartphone, whether that’s from home or in the office. One broker said he got a quote while sat in bed – we didn’t ask what time of day.”

Hearsum says the plan, over time, is to directly integrate with brokers’ own electronic platforms.

“What we have is still unique, so brokers aren’t feeling fatigued by it, but rather the opposite. That said, we don’t see the platform as the long-term vision of Ki. We see future as direct integration with broker systems. We will be announcing a broker integration very shortly, and we’re having conversations with several brokers on similar integrations,” he says.

No buttering up

While speed and simplicity for quotes is one advantage, data quality is also essential to feed the algorithm. The more data the algorithm gets, the better it becomes.

“It’s critical that we have consistency in the data. There is already remarkable consistency between brokers, which is fabulous,” Hearsum says. “The algorithm thrives on data, data and more data. An underwriter can look at a few screens at once, but the algorithm has no such limitations. The more data we can get from the broker, the more sophisticated and powerful our algorithm can become.”

With no human interaction from Ki’s side in the quoting process, human bias, whether conscious or unconscious, is removed as a risk from making underwriting decisions, Hearsum explains.

“There are no favoured or unfavoured brokers and there’s no need to butter up the algorithm. We’re finding that brokers value absolute consistency with Ki. They like that it’s so clear cut. There is no offering a deal in order to get another, better one later. Consistency is critical to how it operates, which you can’t say the same for human beings,” he says.

“A broker may be disappointed that they are offered $1.5m of capacity when really they want $3m. Another time it will offer them the full amount and they’re very happy. There’s no negotiating process with the algorithm, only the data, and it will always offer its maximum line,” he adds.