Australia’s general insurance sector is absorbing the consequences of a structurally worsening natural disaster cycle. Against that backdrop, IAG’s decision to deploy customer-facing AI voice agents directly into its claims operations – rather than in back-office or staff-facing functions – marks a shift in how the sector applies artificial intelligence, and one with governance and conduct implications that extend well beyond the announcement itself.
IAG confirmed on July 27 it will partner with OpenAI to develop agentic voice solutions for its Australian operations using OpenAI Presence, a newly launched agentic system. The initial focus is natural perils claims, where demand on frontline staff spikes rapidly and response consistency matters most. Delivery is expected in the first half of FY27.
Extreme weather events generated almost $3.5 billion in insured losses from 264,000 claims across Australia in 2025. The costliest single event was Ex-Tropical Cyclone Alfred, which produced 132,000 claims and more than $1.5 billion in insured losses, while two severe storm events in October and November together generated a combined $1.4 billion in losses. Munich Re’s NatCatSERVICE global database shows that for most of the last 45 years, Australia has ranked second only to the US for economic and insured losses per capita. For a general insurer of IAG’s scale, that exposure trajectory makes natural perils claims volume a structural operational problem, not a cyclical one.
IAG CEO Retail Insurance Australia Julie Batch said the decision to focus on natural perils was deliberate. “Our initial focus will be on where the need is greatest – high volume natural perils – to ensure we help our customers the best way we can at a stressful point in their lives,” Batch said. Satya Tammreddy, OpenAI’s head of go-to-market for Australia and New Zealand, said the aim was to expand service capacity and reduce the operational burden on frontline staff. “We look forward to working with IAG to explore how customer agents can help resolve issues faster, expand service coverage, and deliver more consistent experiences, while reducing the operational burden on frontline teams,” Tammreddy said.
The announcement lands as other major Australian general insurers accelerate their own agentic AI programs. Suncorp CEO Steve Johnston confirmed during the company’s half-year FY26 earnings call that its agentic program is in full-scale delivery, with initial deployments across claims and customer services processes on track. A centrepiece of Suncorp’s existing AI stack is the Single View of Claim (SVOC) tool, which consolidates case notes, customer communications, and supporting documents into a unified summary and recommends next steps. Suncorp says the tool reduces claim review time by between five and 30 minutes, depending on complexity.
Allianz also deployed agentic AI in Australian claims in 2025. Its Project Nemo system, launched in July 2025, uses seven specialised AI agents to automate low-complexity food spoilage claims and reported an 80% reduction in claim processing and settlement time. Importantly, Allianz’s model retains a human professional as the final decision-maker on payouts – a governance design choice that sets a visible benchmark for how agentic claims automation can be structured while maintaining human accountability.
The distinction between these deployments and IAG’s announced approach is material. Both Suncorp’s current tools and Allianz’s Project Nemo are oriented toward internal staff support or back-end automation. IAG’s stated application is customer-facing voice interaction – a more direct point of contact, and one carrying greater conduct risk exposure.
The Australian Securities and Investments Commission’s (ASIC) Key Issues Outlook for 2026 named two concerns that intersect directly with IAG’s announcement. ASIC identified agentic AI as a source of potential consumer harm, noting that while it “can help people shop around for deals and avoid loyalty penalties, it can also compound risk given its capability to independently plan and act.” In the same document, ASIC separately flagged poor insurance claims handling following extreme weather events as a 2026 enforcement priority, noting it has commenced court proceedings over serious claims handling failures and that the volumes and costs from disasters highlight ongoing challenges insurers face. Those two threads – agentic AI risk and disaster claims conduct – converge precisely on the use case IAG has announced.
The Australian Prudential Regulation Authority’s (APRA) April 2026 letter to all regulated entities observed that insurers are already moving into customer-facing AI applications including claims triage and customer interaction, and that governance has not matured at the same pace, with boards still developing the technical literacy required to provide effective challenge on AI-related risks. APRA also flagged specific concerns about AI agents as non-human actors: identity and access management capabilities have not yet adjusted to non-human actors such as AI agents, and gaps were observed in security testing for AI implementations. The regulator stated it will take stronger supervisory action, including enforcement, where entities fail to manage AI risks proportionate to their size and complexity.
IAG said its solution will be developed in line with its internal standards for governance, security, and responsible AI use – framing that is consistent with the expectations both regulators have articulated publicly in recent months.
IAG has not disclosed how OpenAI Presence integrates with its existing claims management infrastructure, what escalation thresholds will apply when customer interactions exceed the system’s scope, how the solution will be audited for compliance with the General Insurance Code of Practice’s claims handling obligations, or what governance structures will govern a newly launched agentic product operating in emotionally sensitive interactions at scale. Those details will matter in a regulatory environment where ASIC and APRA have both made clear that governance failures in AI – not AI adoption itself – are where their enforcement attention will be directed. IAG’s first-half FY27 delivery window gives the market a concrete timeline to watch.
