Re/insurers may soon look to artificial intelligence to handle routine work, as bottlenecks leave some business they want to write without a quote, according to new research.

The study, commissioned by mea Platform and conducted by ISG, found that 83 per cent of global insurance-market respondents would support AI carrying out repeatable tasks.

However, mea Platform, an insurance technology company, and ISG, a research and advisory company, found that respondents said they would not hand that work to a general-purpose AI model.

Seventy-five per cent would only allow AI to execute work through a model designed for insurance or governed by the company’s own rules, while only 6 per cent would trust a general-purpose model alone for high-consequence underwriting and claims decisions.

The research covered senior insurance leaders in underwriting, claims, operations, technology and transformation across North America, Europe and Asia.

It found carriers estimate that about one in nine broker submissions is declined or left unquoted because they cannot keep pace, even where they have an appetite for the risk.

Martin Henley, chief executive of mea Platform, said: “Carriers are turning away business they had the appetite to write because operations cannot keep up, and winning that back is worth more than any saving.”

Among firms already using AI in operations, 61 per cent reported productivity improvements and 51 per cent reported faster turnaround times. Respondents expected operating costs to fall by an average of 16 per cent over the next two years.

The report said insurers see the main value is in automating the most repetitive processes like submission intake, triage, quote generation, bordereaux processing, claims handling and compliance screening. Human staff would still make the most important decisions.

Ninety-six per cent of respondents said AI-led redesign was on their agenda. Yet only 13 per cent have reached an advanced operating model where AI is central to workflow design, while fewer than 1 per cent use fully AI-native processes today.