Automated claims systems speed processing at major German carriers but spark complaints over wrongful denials

Automated claims systems speed processing at major German carriers but spark complaints over wrongful denials

By [Staff Writer]

FRANKFURT — Major German insurers have accelerated the use of automated claims‑processing and AI‑driven fraud detection systems to cut cycle times and investigative costs, but consumer advocates, regulators and ombudspersons say the technology has also increased the risk of wrongful denials and opaque decisions that leave policyholders without timely payouts. Insurers and vendors say automation speeds settlements and reduces fraud; regulators and consumer groups say data quality, model errors and a lack of transparent individual explanations are driving more complaints and raising legal and supervisory questions. (Who: major German carriers, vendors, regulators, consumer groups. What: expansion of automated claims and fraud detection; complaints about wrongful denials. When: deployment accelerated since 2023–2024 and intensifying through 2025–2026. Where: Germany and, by example, other advanced markets. Why: automation reduces costs and speeds handling but introduces false positives, biased outputs and accountability gaps.) (eiopa.europa.eu)

What changed — and why insurers embraced automation
Insurers across Europe and in Germany have pushed automation across the entire claims chain — from initial intake and document parsing to automated damage estimates, straight‑through payments for simple losses and AI risk scoring to flag suspected fraud for investigation. Industry and regulator surveys show rapid adoption: EIOPA’s market work and industry reporting indicate roughly half of non‑life insurers were already using AI across the value chain by 2024, and many more expanded deployments in 2025. Insurers cite faster settlements, lower administrative costs and improved detection of organised fraud rings as primary drivers. (eiopa.europa.eu)

Carriers and insurtech partners publish prominent efficiency metrics. Vendors and some insurers report dramatic reductions in handling times and increases in “straight‑through” processing: examples cited in industry materials include insurers or service partners cutting document processing and damage‑assessment times from days to minutes and raising the share of claims settled automatically. Individual vendors also claim significant reductions in false positives after model tuning and data enrichment. Those performance claims are the basis for broad, continued rollout across German carriers. (globenewswire.com)

Benefits documented — and promoted
Executives argue the business case is clear. Generali, for example, describes automating document extraction, initial damage assessment and routine settlements as central to modernising operations; the company’s public materials note millions of documents processed and wide use of automation in claims and customer communications. AXA Partners, Allianz units and others have announced partnerships with insurtech vendors to price and value small household or vehicle losses automatically, promising faster assistance for customers. Insurer and vendor materials emphasise that the combination of large historical claims datasets, computer vision for photos and video, and real‑time market data allows objective, repeatable decisions that benefit honest claimants and protect premiums. (generali.com)

Where automation runs into trouble: complaints and rising dispute volumes
At the same time, complaints to Germany’s ombudsman and consumer‑advice groups have risen. The independent Versicherungsombudsmann reported record complaint volumes in 2024 and a further surge to roughly 28,000 filings in 2025, with consumers increasingly citing delayed handling, poor explanations and unexplained denials among the main grievances. The ombudsperson, Sibylle Kessal‑Wulf, told the ombudsman’s press conference that 2024 had produced the highest number of complaints since the body’s founding and that many complainants point to slow or opaque insurer communication. (handelsblatt.com)

Regulators and consumer advocates warn that automation raises particular risks when systems are used to triage or decide claims without clear human oversight. BaFin — Germany’s financial regulator — and Europe‑wide authorities have made automation and “dark processing” a supervisory priority, highlighting risks from opaque models, data biases and overreliance on third‑party cloud and model providers. BaFin’s public commentaries stress the need to prevent unjustified discrimination and to embed AI governance into existing ICT and outsourcing rules. In a public warning reported by the press, BaFin’s top supervisor Julia Wiens said insurers must process legitimate claims “zügig” and warned the authority will intervene where handling is unduly delayed. (bafin.de)

Why wrongful denials happen: false positives, data and model issues
Experts and practitioner surveys identify three recurring technical causes of wrongful denials and excessive flags for investigation:

• Data quality and scope: models trained on incomplete, biased or unrepresentative historical claims can misclassify legitimate claims as suspicious. Fraud‑detection vendors and surveys repeatedly rank “internal data quality” and “access to external data” among top implementation challenges. (friss.com)

• Thresholding and operational design: many systems produce risk scores rather than binary decisions; trouble arises when business rules convert those scores into automatic denials, or when high‑risk thresholds push claims into lengthy manual investigations that effectively delay payouts. Vendor marketing stresses “reduced false positives,” but insurers’ operational configurations vary widely. (friss.com)

• Model drift and new fraud modalities: fraud patterns evolve, particularly with generative AI tools now able to fabricate realistic images and documents. Detection models need continuous retraining and human review to avoid misclassifying novel legitimate evidence or failing to catch sophisticated schemes. Recent academic work and industry reports flag the rising challenge of deepfake‑assisted fraud and the need for multi‑layered detection strategies. (nature.com)

Industry surveys confirm that high false‑positive rates are a meaningful implementation problem. In FRISS’s 2024 industry survey, insurers repeatedly listed “a high number of false positives” among the top barriers when deploying fraud‑detection software; the German subset specifically cited false positives as a leading concern. Vendor literature and case studies publish large percentage reductions in false positives after tuning, but those are company claims that depend on customer data and configuration. (friss.com)

Concrete consequences for policyholders
For individual customers, the practical impact of false positives is immediate and material: delayed payments that strain household budgets, requests for additional documentation that are onerous or impossible to produce quickly, and in a minority of cases, outright denials without comprehensible reasons. Consumer groups say this is not a theoretical problem. Klaus Müller, president of Germany’s Verbraucherzentrale Bundesverband (vzbv), warns that algorithmic decisions are often opaque and that consumers lack sufficient rights to a meaningful, case‑specific explanation. He and other consumer advocates call for stronger transparency and easier complaint channels. (vzbv.de)

Cases in other markets illustrate the stakes. In the United States, major health insurers have faced litigation and public scrutiny for allegedly relying on automated systems to deny care, prompting providers and patient advocates to deploy their own AI tools to challenge denials. Those disputes, while in a different regulatory environment, underscore how fast, algorithmic adjudication can escalate into large‑scale consumer harm and legal action if not carefully governed. German and European regulators watch those examples closely as automation expands. (theguardian.com)

Insurers’ defence: human‑in‑the‑loop, audit trails and appeals
Insurer statements and vendor materials emphasise that modern fraud detection is intended to help human investigators prioritise work rather than to substitute for human judgement entirely. Vendors advertise “contextual guidance” for investigators, and several insurers say they preserve human review for higher‑risk cases. Generali and AXA public materials describe training staff and implementing governance frameworks so that automation augments — not replaces — human decisions. But implementation varies across carriers, and consumer advocates say that “human review” is sometimes limited to rapid cursory checks that do not correct a wrongful automated outcome. (generali.com)

Regulatory and supervisory moves
The European Artificial Intelligence Act, published in mid‑2024 and entering into force in August 2024, took a risk‑based approach and created new obligations for “high‑risk” systems; many of its enforcement elements roll out over 2025–2026 and beyond. European insurance supervisors and BaFin have since published guidance and consultations to align the sector with the AI Act and with financial ICT rules, urging insurers to document models, assess fairness, manage third‑party risks and ensure explainability and human oversight where decisions affect rights and livelihoods. EIOPA’s work on “supervising digital transformation” and BaFin’s non‑binding guidance stress that regulators expect objective safeguards and will increase scrutiny where automation intersects with consumer protection and financial stability. (bafin.de)

Enforcement examples and legal pressure
Supervisory bodies have begun to act where transparency and consumer rights are implicated. In Germany, data‑protection authorities have already fined at least one financial institution for failing to provide adequate transparency about an automated decision in an individual case. That enforcement trend — combined with EU jurisprudence and ombudsman activity — signals that insurers may face both supervisory measures and consumer redress if automated processes lack adequate individual explanations and appeals procedures. Consumer organisations and ombudsmen are urging regulators to require insurers to produce case‑level explanations of why a claim was rejected and to make appeals easier. (bitkom-consult.de)

Experts’ prescriptions to lower wrongful denials
Independent researchers, regulator guidance and industry bodies converge on a set of pragmatic controls to reduce false positives and wrongful denials:

• Human‑in‑the‑loop for high‑impact decisions: keep human investigators accountable for denials and ensure those investigators have access to interpretable model outputs and supporting evidence. (eiopa.europa.eu)

• Better data governance: improve internal data quality, enrich training sets with representative examples, and share anonymised fraud‑pattern information across the market where lawful, to reduce model blind spots. Industry surveys identify data quality as the single biggest implementation barrier. (friss.com)

• Explainability and case‑level disclosure: require operators to provide individualised explanations when a claim is negatively affected by algorithmic scoring, in line with data‑protection principles and emerging case law. European consumer advocates stress this as essential to practical contestability. (vzbv.de)

• Continuous monitoring and independent audits: test for bias, track false‑positive and false‑negative rates over time, and subject models and vendor pipelines to third‑party audits. EIOPA and BaFin guidance call for integrating AI risk management into existing ICT and third‑party oversight frameworks. (eiopa.europa.eu)

• Clear appeals and speedy processing: regulators in Germany have urged insurers to process legitimate claims within reasonable timelines and to streamline remedy and appeal mechanisms; ombudsman data show growing dispute volumes where those expectations are not met. (handelsblatt.com)

Voices from the field
“We expect insurers to work quickly. Mangelnde Personalressourcen oder ein erhöhtes Schadenaufkommen können kein Grund für dauerhafte Verzögerungen in der Bearbeitung von Leistungsanträgen sein,” Julia Wiens, a senior official at BaFin, told reporters when urging faster handling and oversight. Consumer advocates echo the point: “Die EU‑Kommission und die Bundesregierung scheinen dem Druck der Wirtschaft nachzugeben. Das ist enttäuschend und kurzsichtig,” said Klaus Müller, president of the German consumer group vzbv, arguing for stronger transparency and quality requirements for algorithmic systems. Industry vendors counter that the new tools make the system fairer and faster and that responsible deployments include human oversight and audit trails. (handelsblatt.com)

The trade‑off: faster payouts and premium protection vs. contestability and trust
Insurance executives and investors see automation as a necessary modernization: faster claims reduce customer friction for straightforward losses and cut costs that otherwise raise premiums for all customers. Regulators and consumer groups see the opposite risk: if automated systems produce opaque, unexplained denials and frequent false positives, public trust in insurers will erode and vulnerable households will suffer avoidable hardship. Both sides therefore point to governance, transparency and independent oversight as the decisive elements that determine whether automation will improve or impair the market. (generali.com)

What to watch next

• Supervisory enforcement: BaFin and national data authorities are increasing scrutiny and may require case‑level transparency or impose governance remediation measures where automated decisions cause systematic consumer harm. Courts and the European Court of Justice are also clarifying the reach of automated‑decision rules under data‑protection law. (jonesday.com)

• Market adjustments: insurers will likely continue piloting automation but must demonstrate lower false‑positive rates, robust human review and clear appeals workflows if they hope to avoid rising complaint volumes and reputational damage. Vendor claims of large reductions in false positives are persuasive in sales materials, but independent validation and peer‑reviewed studies remain limited. (friss.com)

• Policy changes: the phased enforcement of the EU AI Act and related supervisory guidance will force insurers to document risk‑management processes and may require stronger transparency and audit capabilities for systems that materially affect consumers. Industry lobbying continues, but regulators stress consumer protection and financial‑stability considerations. (theguardian.com)

Bottom line
Automated claims systems have demonstrable efficiency and fraud‑detection benefits that major German carriers and their technology partners promote aggressively. But the rapid roll‑out has surfaced predictable technology risks — data quality, model bias and operational choices that convert risk scores into automatic denials — and those risks have translated into more consumer complaints and regulatory attention. The path forward will depend less on whether insurers adopt automation than on how they govern it: transparent, auditable models, meaningful human oversight, clear individual explanations and faster remedies for mistaken denials. In the absence of those safeguards, automation’s gains in speed and cost could be eclipsed by rising disputes, enforcement actions and erosion of public trust. (handelsblatt.com)

Sources and further reporting
This article draws on ombudsman and industry reports, regulator guidance, vendor white papers and academic studies, including the Versicherungsombudsmann’s activity reports and press statements (2024–2025), BaFin and EIOPA guidance on AI in financial services, vendor and insurtech materials from FRISS, Shift Technology and others, and recent academic literature on fraud detection and model performance. Key sources: Versicherungsombudsmann e.V.; BaFin; European Insurance and Occupational Pensions Authority (EIOPA); Verbraucherzentrale Bundesverband (vzbv); FRISS Insurance Fraud Report 2024; Shift Technology and vendor releases; recent peer‑reviewed studies on fraud detection and anomaly detection. (versicherungsjournal.de)

(Reporting contributed by [staff/analysts]; interviews and document review conducted January–February 2026.)