Two of Europe's largest claims operations are now settling the bulk of claims in under one hour, end-to-end, in production. Autonomous agents execute the full process, and human experts handle what agents are built to surface: the cases that genuinely require judgment, negotiation, or regulatory discretion. Everything else runs itself.
Both CED and Allianz Partners made this shift on existing infrastructure, without replacing core systems or expanding their operations teams. The results are in production, not in a controlled pilot.
CED: 600,000 Claims Per Year, Zero Backlog
CED is a third-party administrator processing over 600,000 claims annually. The high-volume, low-exception work that consumes most of an operations team's time across classification, triage, and routing now runs autonomously. Backlogs have been eliminated.
What makes this significant isn't the speed. It's what the speed reflects. Francois Goffinet, CEO of CED, described the organizational logic behind the deployment:
"We need to remove the administrative burden away from our core resources to make sure we focus them on what really matters and what really creates value for the end customer."
When agents absorb the processing volume, experienced claims professionals stop spending their days on classification work and start applying their expertise where it produces results: on complex cases, customer escalations, the work that actually requires a human in the loop.
Allianz Partners: Governed Autonomy Across Five Countries
Allianz Partners is rolling out claims processing end-to-end across 30+ countries and multiple lines of business within its portfolio, all on existing infrastructure. For a global insurer operating under regulatory scrutiny across multiple jurisdictions, the operational challenge isn't just throughput. It's consistency. The same claim type handled differently in two markets is both a compliance exposure and a customer experience failure.
Pieter Viljoen, Chief Data Officer of Allianz Partners, framed what autonomous operations address at that scale:
"It's about consistency, accuracy, and quality of decision-making, as well as significantly lifting the work of operations."
When agents execute under governed rules, every market produces the same decision quality by design. Variance doesn't require manual oversight to control. It's engineered out at the model level, which is a fundamentally different relationship between governance and operations than traditional automation provides.
What This Operating Model Actually Looks Like
During a live demonstration at Otera's Autonomous event, a travel insurance claim moved from initial submission through validation, completeness checks, policy verification, and final settlement without human intervention at any stage. The agents handled the full range of unstructured inputs, including emails, scanned documents, and photographs, ran the relevant policy logic, and produced a settlement decision in under 90 seconds.
The distinction worth drawing here is not about speed. It's about where human judgment sits in the process. Workflow tools and RPA digitize individual steps, but humans still bridge systems, reconcile data across handoffs, and make judgment calls wherever the process produces ambiguity. In the model CED and Allianz Partners are running, agents own the process end-to-end. Humans own the governance layer and intervene when the agent calls for it, not as a structural requirement of the process itself.
What Comes Next
The question of whether autonomous operations are viable in regulated enterprise environments has been answered by what's already running in production. The more useful question now is an execution one: what does it actually take for an insurer to move from a staffed manual process to an agent-run one, and where do organizations typically get stuck in that transition?
That's the conversation worth having. To see exactly how CED and Allianz Partners got there, get the blueprint.
Next Steps: From Insight to Action
Next Steps: From Insight to Action
.svg){kind=icon}
