Rethinking AI in Insurance: It’s Time We Asked the Right Question

display-middle-ads-here

The insurance industry, like many others, is racing to adopt artificial intelligence. Yet amid the excitement, the real question often gets lost: what exactly are we trying to achieve?

In countless research conversations, I’ve heard financial services leaders declare that AI is the future. It’s the buzzword at conferences, the shiny object in boardroom strategies, and a centerpiece of many vendor pitches. 

But often, that enthusiasm isn’t grounded in tangible purpose.

If you ask senior risk managers within the financial sector what business problem AI is solving or what measurable value it will deliver, the answers can often be vague, and it feels like there’s a disconnect between ambition and execution that we urgently need to close.

The truth is, implementing AI without a clear objective is like trying to build a house without blueprints: you might assemble something impressive looking, but it likely won’t stand the test of time.

To create real business value, insurers must define the use case upfront – what’s the goal, what will success look like, and how will it tie back to the organization’s core strategy? 

AI should be used to solve specific problems: reducing operational costs, improving customer experience, speeding up claims processing, or enhancing risk selection. If those connections aren’t explicit, we risk investing in technology for technology’s sake.

One of the biggest challenges is that our data – the lifeblood of any AI solution – isn’t in good shape – a concept that won’t come as a surprise to anyone who’s worked in insurance for more than a week.

Most data is incomplete, inconsistent, or siloed across functions. Underwriters don’t always have access to claims data that could improve risk selection, and catastrophe modelling processes often don’t feed back into underwriting decisions.

This lack of circular connection between departments creates blind spots, and poor data input inevitably leads to poor AI output.

What comes as a result is the temptation is to charge ahead with advanced AI capabilities, hoping they’ll transform our business processes.

But before we get our hands on that prize, we have to put in some hard yards on the data sources that we use. The priority shouldn’t just be the implementation of AI – it should also be about using AI to fix the data first. 

I’ve seen AI used effectively to automate data reconciliation, create intelligent matching logic, and enrich missing fields by referencing external sources. These aren’t glamorous projects, but they create the foundation on which everything else can be built.

That foundational work also forces us to think beyond our individual silos. One of the most dangerous habits in insurance is building processes and systems that serve only one part of the business.