Underwriting has always been an exercise in incomplete information: a broker’s submission, an engineer’s report, a loss history that may be thin or inconsistent and a pricing model that is only as good as the data fed into it. What’s changed - rapidly - is the industry’s confidence that machines can help close those gaps at scale. Across Australia and New Zealand, AI is increasingly being positioned not as a single tool but as a new layer of capability that alters how insurers manage risk, learn from claims and run underwriting operations.
That shift comes through most clearly when underwriters describe AI not in abstract terms, but as a practical way to improve decisions and intervene earlier. Eric Lowenstein (pictured), CEO of healthcare underwriting agency Tego, said one major change is AI’s ability to deepen the “risk conversation” with insureds - moving beyond the binary of accept/decline and into a more preventive posture.
“It allows us to look at risks sometimes more favourably and allows us to get a better understanding of the exposure," said Lowenstein. "Equally, it can sometimes provide us deeper insights which may impact how we want to address the risk going forward or how we want to educate or advise the insured on how to improve their policy and procedure.”
In lay terms, AI’s first underwriting transformation is this: risk management stops being a once-a-year renewal event for underwriters and brokers and becomes a continuous loop. If models can draw insight from richer datasets that include incident trends, controls, procedures, even operational signals - then underwriting can start to look more like prevention than post-loss funding.
Watch next: Will artificial intelligence take brokers’ jobs?
This can have major implications for brokers, particularly in nat cat risk prone countries like Australia and New Zealand where insurers are already being pushed toward more active risk selection and risk improvement by rising building and repair costs, catastrophe volatility, supply-chain fragility and greater scrutiny of governance and duty of care. AI can support better segmentation (who is actually safer), but it also raises a new question for underwriters and brokers: if you can see risk signals earlier, what are your expectations to act on them? The industry’s risk “baseline” may lift - clients could face more specific control requirements, more structured risk questionnaires and more evidence-based endorsements.
There’s also a counterweight: AI can create new concentrations and new risks. The Reserve Bank of New Zealand has flagged both the potential benefits of AI (including better risk management and productivity) and the potential risks, including reliance on a small number of third-party providers and the chance that errors or complexity amplify existing vulnerabilities. In underwriting terms, that is a warning against blindly outsourcing judgement to black-box tools - particularly where the consequences of systematic error could be widespread.
The second transformation is the way AI can compress the distance between what happens in claims and what changes in underwriting. In traditional insurance cycles, claims insights can take months to show up in actuarial analysis, committee papers and revised guidelines. AI promises a faster loop.
“On the claims side it’s being able to look at trends in real time so that you can respond to risk and respond to needs in real time," said Lowenstein.
For underwriters, “real time” doesn’t necessarily mean live pricing changes by the hour. It can mean earlier detection of loss patterns - like specific injury types, failure modes, fraud vectors, or third-party litigation trends) and earlier intervention - through risk advice, revised appetite, tighter endorsements, or claims triage changes that reduce leakage.
Australian market data points to why this is more than theory. Gallagher Bassett’s Carrier Perspective: 2025 Claims Insights whitepaper found that 88% of Australian insurers reported using generative AI in some part of claims resolution, with use cases such as intake, triage, fraud detection and customer communications. Even if much of this is still task-based automation, it signals that claims workflows are quickly becoming AI-enabled - and that underwriting will increasingly inherit the outputs including cleaner narratives, categorised causes of loss, structured documents and faster trend reporting.
A second Australian example is the Insurance Council of Australia’s (ICA's) announcement of a collaboration with Shift Technology and EXL to build a national data analytics fraud detection and investigations platform - starting with motor claims and intended to alert investigators to suspicious activity in real time. For underwriters and brokers, claims behaviour and fraud patterns will be identified sooner and appetite or pricing responses may follow more quickly, especially in pressured personal and SME portfolios.
The third transformation is less visible to customers but potentially the most disruptive to market structure: AI as a data-quality revolution that overcomes the underwriter's daily struggle with unstructured, fragmented and inconsistent information. If underwriting can ingest messy submissions, extract renewal data, standardise it and compare it across portfolios, then speed and consistency become strategic advantages.
“This is about being able to extract data a lot smarter, being able to aggregate and use big data in a way that we've never been able to do before, because sometimes the historical way of gathering renewal information can be from multiple sources including different types of forms and different reports," said Lowenstein. The ability to aggregate all of that into an AI platform and compare it with other insureds and then also predict future risk and pricing is a big enhancement for underwriters.
For brokers, this can cut two ways. On one hand, faster turnaround and clearer questions can improve client experience. On the other, it may raise the bar on submission quality - and shift negotiation dynamics if underwriters feel they have stronger, faster evidence for pricing or coverage positions.
So far, AI in underwriting is still an evolving issue and deployment among underwriters is very uneven. A 2025 Bain & Company claims maturity assessment report suggested many insurers globally remain in early-stage experimentation, with only a small fraction having scaled AI broadly. That gap - between pilots and enterprise adoption - is where many Australia and New Zealand underwriters and brokers are operating now. Those challenges include dealing with the near-term friction of change while trying to capture the long-term benefits of AI that revolve around improved risk management, data quality and claims learnings.
