By Manjit Rana, Executive VP Insurance at Clearspeed
Insurance, at its core, is a business of trust.
Every insurance agreement rests on the assumption that the information entering the system is honest and accurate. And, historically, that hasn’t been a problem. Faking evidence took effort, and the cost of an occasional bad actor could be priced into a carrier’s books.
Today, that’s no longer the case. With generative AI, a doctored invoice or photorealistic image of damages can be produced in seconds – and that’s exactly what insurers are seeing.
According to Verisk, 98% of insurers say AI is fuelling media fraud, yet just 32% are confident they could detect a deepfake. And, alarmingly, it’s becoming more culturally permissible to manipulate information. The same research found that 55% of Gen Z and 49% of Millennials would consider a small rule-bending edit to a claim photo.
As a result, the trust deficit is growing, driving up loss ratios. Indeed, the Coalition Against Insurance Fraud estimates that fraud costs the industry $308.6 billion every year.
The issue with the guilty until proven innocent model
A big part of the problem lies in the fact that insurers have historically lacked a consistent way to trust the genuine majority quickly, which doesn’t enable them to focus human judgement on those cases that need it.
For 30 years, the industry has invested heavily in trying to detect the minority of bad actors, which has resulted in processes that subject every claimant to similar levels of scrutiny before genuine claims can be distinguished from potentially fraudulent ones. It’s a slow process that involves asking and double checking everyone, meaning everyone suffers the same frictions before the genuine majority are separated from the dishonest few.
And that model is equally now buckling as fraud gets easier and claims volumes rise. While existing fraud models flag between 30-50% of claims, special investigation units can typically only work on 2-5% of cases.
Clearly, it’s a problem for insurers, but also their insureds. If just 5% of claims are actually fraudulent, up to half go through a slow, laborious claims process, it’s the honest majority who pay with a poor customer experience. How many of the 95 genuine claimants in every 100 come away feeling distrusted because the process was designed to catch the other five?
It’s likely because of this that only 37% of Americans believe their insurer will help them with problems that arise. Meanwhile, J.D. Power’s 2024 research puts just 15% of customers in the high-trust band, and more than half (51%) in the low-trust band.
So, neither side wins, with both insurers and policyholders suffering from friction, delays and operational drag.
Are insurers automating faster than they can verify?
The crux of the issue is that insurers still lack a consistent way of quickly determining which cases are clear, and which require greater scrutiny or human judgement.
At a time in which automation is becoming increasingly prevalent in insurer processes, that is a problem. Indeed, one of the major questions facing the insurance industry as AI adoption accelerates is whether decisions are being automated faster than the infrastructure needed to verify the information behind them.
Speed without proper judgement isn’t efficiency, but exposure. The more autonomous insurance becomes, the more its economics depend on a capability the industry has never formally built: a way to establish trust in real time, at scale, every time a policy is priced, changed, renewed, claimed against and paid.
The answer to this challenge is making trust a measurable operating layer across the policyholder journey.
Enter the trust intelligence layer: a regulatory-ready risk signal that indicates to decision-makers which interactions need a closer look. This doesn’t replace people – the signals are contributory, providing better direction. But the insurance professional still makes the final call.
Claims seems like the obvious place for such a technology, and is an area where it’s already proving its value. However, trust intelligence isn’t a single claims checkpoint, but a methodology that can be applied across the insurance lifecycle, from policy application to renewals to first notice of loss (FNOL) and settlement.
The benefits of this approach are multiple.
Trust becomes bi-directional. Not only is the focus on ensuring that the carrier can trust the input it prices or pays on, but also on ensuring that the policyholder can trust that the insurer is working in their interests.
Equally, the emphasis is shifted from catching bad actors to clearing honest customers. Genuine policyholders aren’t treated as suspects, which helps address operational drag and provides a better customer experience.
And trust is established from the off. As one former group claims executive told us, “to see claims as the way to reduce fraud is shutting the stable door after the horse has bolted.” In other words, the cheapest way to establish trust is doing so at the first moment of contact – at application or renewal.
Making trust a lasting advantage
For insurers, loss-cost pressure, manipulated evidence, automation and AI governance scrutiny are collectively posing a critical question: can they trust the information their systems are acting on?
Establishing a trust intelligence layer gives assurance in that answer.
For claims teams, it improves speed and focus. For underwriters, it strengthens confidence at policy purchase and renewal. For risk, data and AI leaders, it provides a signal that is auditable and defensible.
The business case is also quantified by metrics that carriers already track such as cycle time, loss-adjustment expenses, loss ratios, and customer retention rates. Clearing genuine claims accurately within minutes removes delays, eases backlogs and cuts costs, while a quick, respectful claims experience keeps customers happy.
The strategic window is open, but it will not stay open for long. In the age of agentic AI, speed on its own will become a liability. The lasting advantage is speed with trust.
