Who chooses your insurance before you do? How AI systems are already shaping the perception of insurers today

Allianz, ERGO, and UNIQA are currently investing in artificial intelligence in very distinct ways - one is scaling across the entire group, another is revamping customer service, and a third is progressively automating core operational business. This raises a question that is rarely asked in executive suites yet: What does an AI answer when a potential customer asks it to recommend a suitable insurer?

The question "Which insurance fits me best?" is increasingly no longer typed into a search engine, but posed directly to an AI system. Whether this is already happening on a significant scale cannot be reliably quantified, as a robust empirical foundation is lacking. However, it is undeniable that the technical development making this question relevant at all has long since become a reality. Insurers are structuring their knowledge for machines, automating core processes, and increasingly deploying generative systems at customer touchpoints.

How an AI system arrives at an answer in the first place

A generative AI system is not a traditional database that retrieves a single, authoritative dataset for every query. Depending on the architecture, it draws on trained knowledge, web searches, or retrieval systems that compile relevant documents from various sources at the time of the query and formulate an answer based on them. This means that if both a press release regarding efficiency gains and a business magazine article concerning job cuts are in circulation about an insurer, it is not "the truth" that decides which narrative dominates the response, but rather the system's weighting of those respective sources.

Precisely at this juncture begins what EYECONIFAY defines as AI Perception Management — a term coined by EYECONIFAY, not an established industry standard.

Allianz: Scaling as a strategy

In January 2026, Allianz SE and Anthropic announced a global partnership focusing on three key areas: Anthropic’s Claude models are being integrated into the internal AI platform free of charge for all employees; "Claude Code" is already utilized by thousands of developers worldwide, according to Allianz; and jointly developed AI agents are set to automate multi-step workflows, such as claims processing in motor and health insurance. Allianz emphasizes a "human-in-the-loop" principle, ensuring that employees continue to handle sensitive or complex cases [1].

Since November 2024, Allianz has supported the voluntary EU "AI Pact," which aims for early alignment with the principles of the EU AI Act [2]. According to its own figures, Allianz has registered more than 900 internal AI use cases group-wide [3]. In June 2026, Allianz also reported that it ranked first among 30 evaluated insurers in the Evident AI Index 2026 — a ranking compiled by the external analytics firm Evident — with an AI talent pool reportedly 28 percent larger than that of its closest competitor [4]. This classification originates from Allianz’s proprietary reporting on the external ranking; an independent verification of the underlying methodology was not conducted within the scope of this article.

From an AIPM perspective, this density of announcements is ambivalent. On the one hand, it creates a consistent, repeatedly validated external image. On the other hand, with every new partnership and every new internal tool, the volume of material from which an AI system must choose also expands — high communication velocity does not automatically equate to high consistency.

ERGO: Two accurate but differently framed narratives

ERGO illustrates how deeply AI is already penetrating the customer interface — and how divergent the public narrative becomes depending on which source is consulted. Chief Digital Officer Mark Klein stated at a press event in June 2026 that the group intends to invest approximately 130 million euros in AI over the next five years to make core processes "better, more sustainable, and more cost-efficient."

According to Klein, the majority of the more than 300 AI and digital projects already deployed are delivering value, though he did not disclose exact figures. Key expansions include process mining (40 processes across ten countries, utilizing Celonis), robotics (660 applications across 13 countries), and voice bots that handle roughly 15,000 calls daily and accurately identify customer intent in about 96 percent of cases — with guaranteed transfer to human agents upon request [5]. When asked, responsible managers could not precisely quantify whether or how strongly the deployment of AI tangibly impacts customer satisfaction (Net Promoter Score) [5].

Concurrently, Labor Director Lena Lindemann confirmed to Bloomberg News in April 2026 a planned reduction of approximately 1,000 jobs by 2030, predominantly in Germany, at a rate of about 200 positions per year. She explicitly emphasized that these reductions do not constitute operational layoffs, but rather rely on voluntary arrangements such as phased retirement and severance programs. Concurrently, ERGO is establishing its own "Reskilling Academy," which up to 700 employees are slated to complete by 2030; approximately 50 individuals were already enrolled as of April 2026 [6].

These two narrative strands — efficiency gains on one hand, socially cushioned but tangible job reductions on the other — do not factually contradict each other, yet they stem from distinct source genres: corporate communications emphasize the former, while business journalism highlights the latter. For an AI system tasked with generating a single, synthesized answer, the critical question is how it weights these inputs — and whether it accurately reproduces or oversimplifies the nuance emphasized by Lindemann herself ("no layoffs," voluntary programs).

AIPM audit applied: ERGO

The EYECONIFAY audit framework — an independent, non-standardized methodology — can be applied concretely to ERGO based on the verified sources cited above:

Are the sources authoritative and unambiguous? Yes — the statements are attributed by name to CDO Mark Klein and Labor Director Lena Lindemann and are documented in established trade media and Bloomberg, respectively.

Is the information consistent across different sources? Partly. The efficiency narrative and the personnel reduction narrative do not contradict each other, but appear across different media with varying emphasis and levels of detail — such as the vital nuance "no layoffs," which is omitted in certain summaries.

Is the information current? Yes, both cited sources date from April and June 2026.

Can initiatives and key actors be clearly distinguished? Yes — statements by Klein (digital strategy, investments) and Lindemann (HR strategy) are assigned to distinct areas of responsibility.

Is the information technically structured for optimal system processing? This cannot be definitively assessed based on available journalistic sources; a separate evaluation of ERGO’s proprietary digital channels would be required.

What do AI systems actually output upon a specific request? There are no proprietary test results available for this; consequently, this article deliberately makes no unsubstantiated claims regarding specific system outputs.

Which of these aspects are commercially material? High: Whether an AI system accurately reproduces the nuance between "job reductions" and "socially responsible, voluntary job reductions without layoffs" can tangibly influence public perception of the company both as an employer and as an insurer.

UNIQA: The regional bridge-building

UNIQA, which maintains a long-standing market presence in Bosnia-Herzegovina, Serbia, and Croatia, demonstrates a third, deliberately sober approach. Operations Board Member Wolf Gerlach publicly emphasized as early as December 2023 a distinction crucial for the accuracy of AI statements: Not every automation constitutes artificial intelligence. For instance, a robotic process automation system that processes seasonal car contract cancellations "does not learn anything" and is therefore technically not classified as AI by UNIQA.

According to company figures, these software robots processed more than 500,000 cases or sub-processes in Austria in 2023, absorbing approximately 11,000 hours of manual labor [7].

For genuine AI applications, UNIQA collaborates with the specialized firm omni:us in the domain of claims automation. According to company data, the share of fully automated motor claims processed in Austria surged from zero to approximately 17 percent within a single year (as of the end of Q3 2023); more than 8,000 comprehensive motor claims were handled entirely by AI in 2023, and over 160,000 benefit claims were processed automatically in health insurance — with the stipulation that claim rejections are never decided solely by AI, but are always subject to final review by human staff. Furthermore, UNIQA leverages the operational capacity freed up by automation to service third-party clients: as early as 2023, company personnel managed claims on behalf of two external insurance firms in Germany [7].

This conceptual precision — distinguishing automation from true AI — extends far beyond mere semantics. An AI system incapable of cleanly separating these two terms would either overstate or understate UNIQA's technological maturity depending on the underlying source. The fact that UNIQA actively communicates this distinction mitigates this risk, though it does not eliminate it entirely as long as third-party sources continue to use the terms interchangeably.

What the three examples share

The three institutions do not reflect a unified strategy, but rather three distinct risk profiles regarding AI-driven perception. At Allianz, the risk lies in the sheer volume of communication, which complicates narrative consistency. At ERGO, it stems from the simultaneous presence of two valid yet divergently framed narratives. At UNIQA, the challenge resides in maintaining conceptual clarity between basic automation and true artificial intelligence. None of these risks has demonstrably materialized thus far, yet all three are structurally inherent, independent of the explicit intent of the respective companies.

What companies can actively influence

No insurer can dictate how ChatGPT, Gemini, or Claude evaluates them. What a company can influence, however, is the quality of its underlying informational architecture: consistent corporate nomenclature across all channels, current and unambiguously attributable information, and verifiable primary sources rather than contradictory secondary reporting. This sounds like traditional corporate communications — and largely, it is. The fundamental distinction lies in the target audience: the message must now be comprehensible not only to human stakeholders, but increasingly to the algorithmic systems tasked with summarizing it.

An established, standardized "AI Reputation Index" or "AI Reputation Score" for insurers does not currently exist, and any figure purporting to provide such a rating without an disclosed, reproducible methodology would be purely promotional assertion. For this exact reason, this article deliberately refrains from offering a quantitative evaluation of Allianz, ERGO, or UNIQA.

Three distinct lessons

Allianz demonstrates that sheer communication volume does not guarantee consistency. ERGO proves that two truthful narratives concerning the same enterprise can generate a fragmented AI profile if essential nuances — such as "no layoffs" — are not consistently maintained across all touchpoints. UNIQA shows that even rigorous internal communication loses efficacy as soon as external sources fail to adopt identical precision. No definitive judgment regarding the actual quality of their current AI perception can be drawn for any of the three companies without systematic auditing of real-world AI outputs. What can be established, however, is a shared mandate: curating one's proprietary digital information base so that algorithms are left with no more room for interpretation than the enterprise itself intends to govern.

Sources [1] Allianz SE, "Allianz and Anthropic Forge Global Partnership to Advance Responsible AI in Insurance," Press Release, January 9, 2026, allianz.com. [2] Allianz, "Allianz joins AI Pact for human-centered AI," November 27, 2024, allianz.com. [3] Allianz, "For People. For Pets. For Trust: Responsible Use of AI at Allianz," March 18, 2026, allianz.com. [4] Allianz, "Allianz ranks #1 in the 2026 Evident AI Index for Insurance, leading 30 global insurers," June 16, 2026, allianz.com. [5] Versicherungsmagazin.de, "Ergo: KI bestimmt die Entwicklung," June 18, 2026. [6] WirtschaftsWoche, "Ergo: Wie die Munich-Re-Tochter mit Jobabbau & Umschulungen auf KI reagiert" (Interview with Lena Lindemann, Author: Bloomberg), April 23, 2026. [7] APA-OTS / UNIQA Insurance Group AG, "Künstliche Intelligenz bei UNIQA," Original Press Release, December 13, 2023, ots.at.

Methodological Note

This article relies exclusively on facts confirmed through direct cross-referencing with original sources. Individual details from earlier research phases that could not be independently verified in the original records (such as an initially reported partnership regarding AI-supported claims inspection or participant figures for an Allianz training program) were omitted or replaced with substantiated data. Statements concerning Allianz, ERGO, and UNIQA are based strictly on the individually audited sources listed above. Where reliable data is unavailable — particularly concerning actual output behavior of AI systems for specific queries — no speculative values or metrics are asserted.

© 2026 EYECONIFAY Research