Allianz
Allianz reports its seven-agent AI cut claim processing and settlement time by 80% in Australia
Allianz launched Project Nemo in Australia in July 2025, a system of seven AI agents that handles food spoilage claims under AUD 500. A human still makes every payout decision. Allianz reports an 80% reduction in claim processing and settlement time, and a launch in under 100 days.
Key figures
Copied exactly from the sources. Copy any figure with its citation.
80%
Reduction in claim processing and settlement time, per Allianz Services executive1
four days or more
Time to process a food spoilage claim before Nemo, per Allianz Technology executive1
one day, or even just hours
Time to process food spoilage claims under AUD 500 with Nemo, per Allianz Technology executive1
seven
Specialised AI agents in the system1
The problem
Severe storms in Australia often cause power cuts, and homeowners come back to spoiled food. These claims are small, under AUD 500, and covered by standard home contents insurance. They also arrive in surges during natural catastrophes, when claims teams focus on bigger losses. According to Allianz, a simple claim could take four days or more.
What they built
Project Nemo is Allianz's first agentic AI system. Agentic AI means several AI programs, each with a defined task, that plan and work through a multi-step process together. Nemo uses seven agents: a planner, a cyber agent that oversees data security, and agents for coverage checks, weather verification, fraud checks and payout calculation. An audit agent reviews the work and writes a summary for a human.
How they did it
Allianz limited the first version to food spoilage claims under AUD 500 in Australia and went live in July 2025. It says the build took under 100 days. A human claims professional reviews the summary and makes the payment decision. Allianz says this is by design and a core principle across its AI uses. The structure is modular, so Allianz is exploring other simple, frequent cases such as travel delays or simple car claims.
Results
An Allianz Services executive says the time to handle and settle claims fell by 80%. Another Allianz executive says food spoilage claims under AUD 500 now take one day or even hours, against several days before. Allianz also says Nemo can process thousands of simple claims in the back office during major catastrophe events, which frees staff for complex or sensitive cases. The company gives no independent audit of these figures.
What to copy
Pick one simple, repetitive claim type, split it into specialised agents, and keep a person accountable for each payout.
Lessons for your project
Start with a narrow, high-volume claim type. Allianz picked low-value food spoilage claims that spike during storms and are simple to judge.
Keep a human on the payment decision. Allianz says payout decisions are never automated; the agents prepare a summary for a claims professional.
Split the work into specialised agents with clear roles, such as coverage, weather, fraud and payout, plus an audit step, so each step can be checked.
Build for reuse. Allianz designed Nemo as modular so it can be adapted to other simple claims, such as travel delays.
Scope the first release narrowly and align the teams. Allianz credits a focused scope and aligned teams for going live in under 100 days.
Caveats
- All figures come from Allianz itself. No independent source confirms the 80% reduction.
- Allianz does not say how the 80% was measured, over what period or sample, or which claims it covers.
- The model or vendor behind the agents is not disclosed.
- The 80% quote refers to claim processing and settlement time. The other timings come from a separate Allianz executive quote and are not precisely defined.
- The article was read through an archived copy because the original page blocked automated access.
- Nemo is described as live from July 2025. Allianz's article is dated 3 November 2025.
Sources
- When the storm clears, so should the claim queue
Allianz · 2025-11-03 · Company blog · allianz.com
Summarized in our own words from the sources above. Spotted an error? See corrections.