Artificial Intelligence at Allianz Two Use Cases

Allianz Group is one of the world’s leading insurers and asset managers serving approximately 125 million private and corporate customers across nearly 70 countries.

In 2024, the company reported total business volume of $208 billion USD and operating profit of $18.5 billion, per an Allianz media release.

Although Allianz does not publicly disclose precise investment amounts dedicated to artificial intelligence, its commitment is evidenced by significant infrastructure development, workforce expansion, and breakthrough deployments of sophisticated AI systems.

The company operates AllianzGPT, an internally hosted generative AI platform launched in September 2023 that serves over 60,000 employees as of early 2025, with deployment targets for all 158,000 employees globally.

Research by Evident AI’s June 2025 Insurance Index identified Allianz as employing approximately 10% of the AI workforce across 30 major North American and European insurers, underscoring deep talent concentration and institutional commitment to AI development.

This analysis focuses on two AI use cases that directly support Allianz’s operational and strategic objectives:

• Automating claims processing for low-complexity events: Deploying agentic AI to reduce claim handling time while maintaining human oversight of payout decisions, thereby improving customer satisfaction and freeing expert claims staff to address complex cases.
• Expediting fraudulent claim management: Deploying supervised learning algorithms trained on historical claims data to instantly flag anomalous claims for human expert review, improvements in fraud detection.

Automating Claims Processing for Low-Complexity Events

Allianz’s traditional claims processing model operated under systematic constraints during natural catastrophe (NatCat) events, such as severe weather disruptions, triggering surges in insurance claims.

These catastrophic events present an operational paradox for insurers: while customers file high-priority claims requiring complex judgment (structural damage, business interruption), they simultaneously generate high-volume, low-complexity claims that consume disproportionate staff attention.

In an Allianz newsletter, Thomas Baach, Managing Director of Core Insurance Platforms at Allianz Technology, characterized the operational constraint as needing to take “four or more days to process” as the focus of the claims teams was on more complex claims happening during the NatCat event.

This operational challenge reflects a broader industry problem. According to McKinsey’s analysis, insurers face significant workflow bottlenecks, legacy process constraints, and customer-experience gaps across the claims lifecycle from traditional claims processes. These pressures and more are leading many to pursue AI-driven solutions to improve operational efficiency.

The global insurance industry processes hundreds of millions of claims annually, with NatCat events creating extreme spikes in claim volume.

According to Allianz’s internal research, and cited by the Insurance Information Institute (III), natural catastrophes accounted for 24% of corporate insurance claims in the United States. Between 2017 and 2021, top loss drivers cost approximately $90 billion, with insurance companies paying over $48 million daily to cover these losses, per the blog from the III linked above.

In July 2025, Allianz launched Project Nemo as a departure from conventional claims automation. Rather than strict use of deterministic rule-based systems that follow predetermined decision trees, Nemo deploys agentic AI — specialized, task-oriented digital agents that independently plan, decide, and collaborate to complete multi-step workflows with minimal human intervention until critical decision points.

For claims related to food spoilage, for example, the system comprises seven specialized agents, per Allianz’s newsletter:

• The Planner Agent orchestrates the entire workflow and maintains process state throughout claim evaluation.
• The Cyber Agent enforces data security protocols and policy guardrails, ensuring no unauthorized access occurs during sensitive claim handling.
• The Coverage Agent verifies whether the claimant’s policy includes coverage for food spoilage caused by severe weather events — the foundational coverage check.
• The Weather Agent integrates with meteorological databases to cross-reference the customer’s stated weather event against recorded conditions, preventing fraudulent claims tied to non-existent weather events.
• The Fraud Agent applies machine learning pattern recognition to identify statistical anomalies suggesting deliberate misrepresentation, flagging high-risk claims for human review.
• The Payout Agent calculates settlement amounts based on claim details, policy terms, limits, and deductibles.
• Finally, the Audit Agent generates a comprehensive summary of all agent decisions and reasoning, creating a complete audit trail for compliance, quality control, and human review.

The entire technical process completes in under five minutes, after which a human claims professional reviews the audit summary and makes the final payout authorization decision.

This architecture embodies what Maria Janssen, Chief Transformation Officer at Allianz Services, describes as Allianz’s core human-based AI governance principle: “With Project Nemo, AI agents support our teams by making recommendations, but the ultimate responsibility always rests with a claims professional.”

Project Nemo launched in Australia in July 2025 and achieved full operational deployment in under 100 days.

According to the Allianz newsletter and Australia’s insurance coverage, Project Nemo produced the following business results:

• 80% reduction in claim processing and settlement time: For eligible food spoilage claims under $327 USD, processing time declined from several days to one day or even hours.
• Near-instantaneous technical processing: The complete seven-agent workflow executes in under five minutes from claim submission to human review readiness.
• Scalability during crisis periods: By automating low-complexity claims, Nemo frees human staff to concentrate on complex cases, dramatically improving response capacity during NatCat events when claim volumes surge and time pressure intensifies.

What distinguishes Project Nemo from a point solution is its modular architecture and explicit design for cross-product scalability. Previously cited press materials note that Allianz is exploring deployment of the agent framework to other low-complexity, high-frequency use cases, including travel delay claims, straightforward auto claims, and property damage assessments.

Expediting Fraudulent Claim Management

By 2023, Allianz’s fraud investigation teams were overwhelmed by growing claim volumes while simultaneously confronting increasingly sophisticated fraud tactics that traditional rule-based systems failed to detect.

The quantitative reality reflected this deterioration, per an Allianz newsletter:

• In 2023, Allianz Commercial detected £77.4 million in claims fraud, up from £70.7 million in 2022.
• Fraud schemes evolved in sophistication and coordination: false Confirmed Claims Experience (CCE) documents (the commercial equivalent of no-claims bonuses) proliferated, suggesting potential ID theft of genuine companies.
• Motor fraud reversed downward trends with a 25% increase in “crash for cash” referrals.

The operational bottleneck was not an absolute lack of detection capability but rather a scarcity of investigative resources. Fraud specialists faced a growing queue of suspicious claims that exceeded their investigative bandwidth, creating delays in processing both legitimate and fraudulent claims.

The crisis extends across the industry, as the Association of British Insurers (ABI) reported that insurers uncovered over 98,400 fraud-related claims in 2024, a 12% increase from 88,100 in 2023. The fundamental problem is detection capacity, not fraud identification methodology, as according to Carpe Data’s 2025 fraud report, traditional fraud detection methods analyze only 5% of open injury claims.

In response, Allianz deployed Incognito in 2023 — a supervised machine learning-based tool designed to identify potentially fraudulent claims.

Allianz doesn’t publicly disclose the system’s architecture or workflow. However, industry standard approaches are extensively documented.

Insurance fraud detection systems typically employ supervised machine learning, where algorithms are trained on historical claims pre-labeled as fraudulent or legitimate, per a study from the International Journal of Advanced Research in Science, Communication and Technology.

According to that study and another study from the Journal of Insurance and Risk, the workflow of industry detection systems operates as follows:

• Claims data are preprocessed; supervised models output fraud probabilities, unsupervised models output anomaly scores per claim.​
• Claims are ranked by these scores, and a top subset is sent to human investigators.​
• Investigators label suspicious cases; these labels serve as ground truth for evaluating model performance.

 Critically, Incognito identifies potentially fraudulent claims, which are then referred to a fraud expert for thorough review and investigation, per an Allianz newsletter. The system does not make fraud determination decisions autonomously. Instead, it functions as an intelligent triage mechanism.

According to Allianz sources cited below, Incognito brought the following business results for the company:

• Allianz UK achieved fraud savings of £37.7 million within the first half of 2024, according to a separate Allianz newsletter.
• Over the first year, overall claim fraud detection increased by 10% from the prior year, per an Allianz newsletter.
• Application fraud savings increased by 150% compared to year-to-date expectations, per an Allianz report.