AI is no longer just a future topic in the insurance industry. This was evident throughout the annual Handelsblatt conference, 'AI in Insurance Summit 2026', held at ERGO Versicherung in Düsseldorf. Many presentations, panel discussions and conversations made it clear that the topic of artificial intelligence has reached the top management of primary insurers and reinsurers. AI is no longer just an experimental niche topic, but a strategic component of future business models, processes, and value creation.
The overall mood was positive, ambitious and forward-looking. Many insurers are already reflecting on concrete progress and are increasingly focusing on how to scale, integrate, and use AI responsibly. However, it was striking that, for an industry whose core business is risk assessment, the risks associated with AI itself received comparatively little attention on the main stage. Topics such as security, control, quality assurance, and testing were only marginally addressed, even though they were a key focus of discussions during breaks.
This is precisely where a key challenge for insurers lies: The more AI is integrated into operational processes, decision-making and customer interactions, the more important robust verification, monitoring and safeguarding mechanisms become.
A common theme running through many of the presentations was the change in terminology. While individual AI use cases have often been the focus in recent years, the Summit increasingly centred on agent-based AI and the orchestration of entire teams of agents. This marks a significant shift in perspective. It is no longer just a matter of using AI to support individual tasks. Instead, the focus is on how AI agents can collaborate across departments and teams, for instance in claims processing, underwriting, customer service, risk assessment or internal control processes.
An important consideration here is that many existing business processes were designed for human workers. They do not automatically align with the way AI agents operate. The industry is therefore beginning to fundamentally rethink processes, adopting an 'AI first' approach rather than treating AI as an after-the-fact extension of existing workflows.
The analysis of the Asian insurance market was particularly impressive. Ping An's example clearly demonstrated how advanced AI-supported workflows already are there. Decisions, such as those relating to claims processing — a process which often takes days or weeks in traditional European insurance — can, in some cases, be communicated to customers within seconds. This insight illustrates the pace made possible by consistently digitised and AI-supported processes.
For primary insurers and reinsurers in Europe, this presents a twofold challenge. Firstly, they must tap into efficiency potential to keep pace with global competition. On the other hand, they must ensure that increased speed does not compromise controllability, traceability and regulatory resilience.
One particularly intriguing point of discussion came from the audience: Can AI risks themselves be insured?
This illustrates how the role of AI in the insurance industry is evolving. AI is not just a tool for process optimisation, automation or decision support. At the same time, AI is becoming a new risk factor that must be assessed, managed and potentially insured.
The discussion revealed that such approaches already exist. This opens up a new field for the industry. Insurers must not only manage AI risks internally, but also understand them as part of their business model.
Regulation and responsibility were common themes throughout the programme, from BaFin’s perspective on regulation in the AI era and approaches to responsible AI, to liability issues in the closing panel. The overall tone was one of reflection rather than caution. The central question is no longer whether AI can be used. What matters is how it is used: responsibly, transparently and for clear benefit.
This responsible approach is particularly relevant for insurers. After all, AI is used in processes involving sensitive data, financial decisions, regulatory requirements and customer expectations. Therefore, it is crucial to evaluate AI systems in terms of not only functionality, but also quality, control, fairness, security and auditability.
One memorable moment at the summit occurred during a presentation on whether AI could be the 'boss of the future'. A participant in the audience spoke up to emphasise her continued assertion of the right to human leadership. This moment was significant because it highlighted an important boundary. Automation can speed up processes, make decisions and reduce employees' workload. However, not every human interaction can be meaningfully replaced by AI. In organisations shaped by trust, experience and responsibility, human interaction remains a core value. AI should therefore not be viewed as a substitute for leadership, communication or responsibility, but rather as a tool that supports and empowers individuals.
In discussions with our customers and partners, as well as behind the scenes at the Summit, it became clear just how prevalent questions about the risks of AI already are. Time and again, the conversation turned to media perception, security incidents, loss of control and how organisations can reliably secure AI systems. This is where our focus topic, 'AI Testing', generated a great deal of interest. After all, as AI systems become more prevalent in business-critical processes, it is insufficient to evaluate their performance just once. Insurers need structured approaches to continuously test AI systems, identify risks and document the results in a traceable manner.
This includes, among other things:
checking AI results for business plausibility,
testing for stability and consistency,
assessing biases and unintended effects,
ensuring the traceability of decisions, and
the integration of AI testing into governance, risk, and compliance structures.
Our experience confirms this: AI testing is a key factor in building trust in AI applications—internally, with regulators, and with customers.
The employees’ perspective also played an important role in many discussions. A works council member at a large insurance company, for example, expressed a nuanced but fundamentally positive view of AI. There, AI was seen less as a means of cutting jobs and more as an opportunity to free up employees from repetitive, routine tasks. This frees up time for higher-value activities such as expert assessments, customer interactions, handling exceptional cases, quality assurance and process improvements.
This perspective is important. After all, a successful AI transformation does not depend solely on the technology. It also depends on employees understanding the benefits, developing trust and being actively involved in the change process.
The AI in Insurance Summit 2026 has shown us impressively that the insurance industry is embracing AI. There is high acceptance and an overall positive sentiment, with a growing focus on scaling, agent-based AI and AI-first processes.
However, as AI becomes more integrated into operational and regulatory processes, it is becoming clear that control (human judgement), testing and traceability are becoming increasingly important. Speed alone is not enough. Only when AI systems are tested, monitored and responsibly embedded into existing governance structures will the foundation for their sustainable use be established.
This presents a key challenge for primary insurers and reinsurers: not only must they introduce AI, they must also ensure it is implemented in a way that allows efficiency, accountability, and trust to work in harmony.
Trustworthy AI requires evidence, not assumptions.
We evaluate AI applications based on technical reliability, security robustness, regulatory resilience and model stability, using measurable KPIs and traceable test results.
Would you like to test AI applications in insurance processes that are secure, traceable and resilient?
Talk to us about how AI testing, quality assurance and governance can work together effectively within your organisation.