Case Study 5 min read
How Allianz Partners used AI agents to boost straight-through claims processing 30% and cut turnaround from days to minutes
Key takeaways
Allianz Partners, a global leader in insurance and assistance services, partnered with Taktile to embed agentic AI across multiple lines of business, starting with its health claims process. By deploying AI on Taktile, which enables experts to easily monitor and collaborate with agents, the Allianz Partners team:
- Lifted straight-through processing of claims by 30% in a production pilot, with direct cost impact
- Improved adjudication quality by 4–6% on selective claims journeys, strengthening fraud and leakage detection
- Reduced claims processing time from days to minutes and improved customer satisfaction significantly
Now the team is expanding the new agentic process to its full portfolio of more than 30 million claims per year.
“Taktile gave us the tools to purposefully build AI into the core of our health claims operations. We’ve gone from a manual, time-consuming process to an agentic workflow that’s already delivering massive value globally.”
Meet Allianz Partners: The global company behind millions of insurance and assistance cases
Allianz Partners is the assistance and insurance arm of Allianz Group, one of the world's largest financial services companies. With 22,200 employees, it delivers travel, health, and automotive coverage to customers worldwide, primarily through partnerships with airlines, banks, and employers who embed Allianz Partners' products into their own offerings. Every year, its teams process tens of millions in cases across a global medical network of 2.1 million providers, making the accuracy and speed of every claim critical to the customer experience.
The opportunity: Build a more efficient health claims process by combining AI agents with expert oversight
The executive team at Allianz Partners recognized a strategic opportunity to evolve the company’s claims operations and address an industry-wide challenge.
Processing thousands of daily claims across diverse languages and formats is an inherently nuanced undertaking. The intricacies of medical documentation often make it difficult for members to submit claims accurately, and the complexity of insurance products has made end-to-end automation challenging to achieve at scale. As a result, reviewers must manage repetitive tasks like manually checking invoices for completeness and confirming basic payment eligibility.
For Allianz Partners, these hurdles presented a clear case for innovation. As one of their senior leaders explains: “Our health claims process was a compelling use case for AI transformation: high volume, high complexity, and significant manual effort.”
Allianz Partners required a solution capable of mastering this organizational and product complexity—and one built for high-stakes, regulated environments.
The vision: combine AI agents with expert oversight to automate and standardize global health claims operations. Once implemented, every health claim could be assessed with the same rigor and accuracy—regardless of region or language—at the speed and scale the business required.
The innovation: Transforming health claims operations with transparent, agentic AI
The team’s goal was to reduce manual tasks like verifying documents and cross-referencing claim details against policies. They needed a solution that could automate time-consuming steps with AI while keeping human experts in the loop on every decision, making every agent action auditable, and ensuring agents follow policy.
First, they worked with Taktile to deploy specialized AI agents as modular components, with each agent responsible for a specific step in the health claims journey.
Operating as a sequential system, these agents handle the most demanding steps throughout the claims journey:
- Document classification – identifying whether an uploaded document is an invoice and ensuring only valid invoices proceed.
- Data extraction – extracting relevant information from invoices (like dates, provider details, currency, and itemized treatments and costs), into a standard format.
- Treatment mapping – categorizing each extracted line item to the correct benefit in the member’s coverage (and flagging any missing coverage).
- Claim assessment – applying the company’s coverage, eligibility, and fraud rules to the claim and invoice details to produce a coverage decision or flag issues for review.
A member of the Allianz Partners management team shares: “Each agent reflects our unique decisioning context – what to extract, how to categorize treatments, and what to flag for review – based on our company policies and guardrails.”
By standardizing data and applying consistent checks, this system enables straight-through processing for clear-cut cases while flagging nuanced issues for expert review. Specialists can follow the logic from the original document through every agent step to the final outcome, providing the full traceability required for rigorous audit and compliance protocols.
This transparency also facilitates a continual expert feedback loop. When the workflow routes high-risk claims to specialists, their insights further refine agent prompts and decision logic where appropriate. The “human in the loop” framework ensures that expert control is maintained across the system, with human oversight governing decisions and feedback at every stage. This enables straight-through processing to increase over time while maintaining clear guardrails, accountability, and human control.
The outcome: Claims processed in minutes instead of days, fewer manual reviews, and a better customer experience
Allianz Partners’ systematic implementation of agentic AI is helping reshape its health claims operations. These wins come from three reinforcing drivers: improved straight-through processing rates, better identification of fraud and leakage, and faster processing time.
An initial 6-month production pilot validated the new agentic workflow bottom-up, with confirmed impact across three key levers:
- 30% increase in straight-through processing of claims, with further gains imminent: More claims now move through the process without manual intervention, and the rate is set to grow as more agents and use cases come online.
- 4–6% improvement in adjudication quality on selective claims journeys: The agentic workflow consistently surfaces fraud and claims leakage that the manual process would miss, applying the same checks to every claim at scale.
- Processing time reduced from days to minutes: Claims that previously took days to process now move through the pipeline in minutes, freeing specialists from repetitive data validation and allowing them to focus on the complex, high-stakes cases that require expert judgment.
As a senior leader at Allianz Partners explains: “The goal was to accelerate and improve how claims are processed, making the experience smoother for both the company and our customers.”
Extending the same agentic foundation across more than 30 million health claims at Allianz will transform what AI-driven operations can deliver at enterprise scale. Long term, this means higher automation, more accurate and consistent decisions, and a better experience for customers, whilst enabling the business to work more efficiently at an enterprise level.
Allianz Partners also onboarded 60+ subject matter experts onto Taktile, empowering them to directly iterate on and backtest complex decision logic. Strict enterprise security and granular access rights allowed them to work quickly within a safe, governed environment, while using Taktile’s advanced simulation capabilities to rapidly test a new hypothesis and deploy to production.
The bigger picture: Integrating agentic AI across the enterprise for long-term, compounding impact
Rebuilding health claims is just the beginning. Because the workflow uses independent, modular agent components, Allianz Partners can reuse these capabilities across the organization without building from scratch.
“By implementing reusable components, we’ve created a foundation for a broader AI transformation program where we can plug-and-play modules into new use cases while keeping traceability and governance consistent.”
