Insurers in Germany that put agents into claims and policy administration need operations managers who can set claim payment limits, staff review, read the workspace and answer to BaFin for the result. Training covers how agents decide and earn depth, DORA's demands on ICT and providers, and the care health data requires. Limits, escalations and pausing an agent stay with the manager.
What does an insurance operations manager in Germany run?
Large, document-heavy teams. Motor, household and liability claims arrive by post, email and broker portals, with repair estimates, photos and police reports attached. Policy administration handles address changes, cover changes, cancellations and the steady flow of letters that German customers still expect on paper.
Private health insurance adds a specialised stream. Policyholders submit doctors' and hospital invoices billed under the medical fee schedule, and the team checks each line against the tariff and the policy before reimbursing. The volume is high and the data is as sensitive as data gets.
The operations manager runs the teams, the service levels and the escalations: complaints, broker queries, cases heading to the insurance ombudsman scheme, and anything a supervisor might later ask about.
Which work can agents take, and at what depth?
- Intake and classification (depth 1). The agent reads incoming post and email, identifies the policy and claim type, and routes it.
- Invoice checking in health claims (depth 1, then 3). The agent compares invoice lines with the fee schedule and policy terms and drafts the reimbursement. After certification, routine reimbursements below the insurer's limit can be paid at depth 3.
- Motor and household claim assessments (depth 2). The agent drafts the assessment from estimates and photos. A claims handler decides.
- Policy changes (depth 3). Address changes and simple cover changes within set rules are processed without a person approving each.
- Claim payments above the limit, refusals and cancellations (depth 4 or held). Held for a named person. A refused claim affects the customer significantly and is always a person's decision.
The operations manager's limits are one of the five bounds every action must meet. The lowest bound wins, and the verdict says which decided.
Will agents replace insurance operations managers?
No. The role shifts towards designing and running a mixed team. The argument is in will AI agents replace operations managers?. An operations manager who understands depth, verdicts and the minimum chain sets limits deliberately; one who does not either leaves agents at depth 1 forever or lets them pay claims nobody has checked.
Review moves up a level too. The manager looks at patterns: held invoices from one hospital, refusals clustering in one product, overrides by one handler. The method is in how operations managers review an AI agent's decisions.
What must the manager learn, and what stays with them?
Learn: the vocabulary of governed agents; how to set payment and processing limits; how to read the workspace (held reasons, certificate status, usage and cost); how to respond to incidents. The agent's certificate lapses after 90 days and a run that clears no depth drops it to depth 1 the same day, so health reimbursements that were flowing will queue until it is re-certified. Managers also need to coach handlers to record corrections well, as in teaching operations managers to bank corrections.
Keep: payment limits and processing rules; the review rota; refusals and complex claims; complaints and ombudsman cases; the decision to pause; and the relationship with the agent provider as an ICT third party. Who does what is easier to agree with a RACI template for AI agent operations.
Systems that record each handler's overrides also measure their work, so the works council will usually want a say. Raise it in the first month, with a clear statement of what review data is for.
Which German and EU rules apply?
BaFin supervises German insurers, within the EU framework coordinated by EIOPA, and expects sound governance and risk management, including over outsourced and automated processes.
DORA (Regulation (EU) 2022/2554) applies to insurers from 17 January 2025. It requires ICT risk management, ICT incident reporting, resilience testing and management of ICT third-party risk. An agent platform is an ICT service, and the operations manager's incident routine feeds the insurer's DORA processes.
The EU AI Act (Regulation (EU) 2024/1689) lists risk assessment and pricing for life and health insurance as high-risk in Annex III. Claims processing is not on that list, but Article 4 requires AI literacy among staff regardless. GDPR governs the data, and health data is special-category data needing additional protection; Article 22 limits decisions based solely on automated processing with significant effects, such as refusing a claim.
This is an explanation of the rules, not legal advice; confirm how they apply to you with counsel.
A 90-day plan
| Weeks | People | Agent |
|---|---|---|
| 1 to 4 | Operations manager, team leads, a works council representative and the data protection officer take Foundations (about 4 hours) | Intake and classification at depth 1 |
| 5 to 8 | Operations manager and leads take Operator modules: reading the workspace, when something goes wrong, usage and billing, running the rhythm | Health invoice checks and claim assessments at depth 2 |
| 9 to 13 | Exam (pass at 32 of 40 questions) and two-week practicum | Routine health reimbursements and policy changes considered for depth 3 |
Copy this into the training plan template for operations managers. For the roles around the manager, see AI agent skills by role.
Where the Academy fits
Pryme Intelligence Academy's Operator track covers what an insurance operations manager does with agents: reading the workspace, keeping knowledge current, when something goes wrong, usage and billing, and running the rhythm. Module tests are set in insurance and in the manager's role. Certified Pryme Operator lasts 12 months, and the admin view tracks expiry across the team.
Questions people ask
Is a claims agent high-risk under the EU AI Act?
Claims processing is not among the Annex III listings, which include risk assessment and pricing for life and health insurance. If the same agent helps price health policies, that use is high-risk.
How does DORA affect an agent incident?
A significant failure of the agent platform may be an ICT incident the insurer must classify and, if major, report. The operations manager's incident log is part of the evidence.
Can the agent refuse a health claim?
It can draft a refusal with reasons. A person should decide, because the effect on the policyholder is significant and GDPR limits solely automated decisions of that kind.