Fifteen requests for proposals, the same data entered seven times, and ultimately a comparison made under time pressure: This is what everyday life looks like for many motor vehicle brokers. Manuel Martin of InsuranceMap wants to shorten this process to just a few minutes. This is achieved using AI that reads policies, solicits quotes, and immediately provides a recommendation.
InsuranceMap, the comparison platform operated by Versicherungsfachmarkt AG, was originally launched as a comparison tool for standard terms and conditions (AVB) in motor vehicle insurance. Today, the platform covers numerous lines of business for private customers: from health insurance to travel insurance and home contents insurance. With the integration of several AI models, InsuranceMap is now taking it a step further: Brokers can upload policies and vehicle documents; the AI automatically generates the quote request, sends it to insurers, collects the responses via chat, and ultimately delivers a comparison ready for the customer, complete with a recommendation. We spoke with Manuel Martin, the platform’s CEO, about the technology behind it, the motivation behind the project, and the future of the tool.
Mr. Martin, InsuranceMap has been around for about six years. What prompted you to take the step toward a fully integrated AI solution now?
InsuranceMap was born out of a specific practical problem. Insurance policies and coverage are extensive and constantly changing. Manually comparing them takes a lot of time in day-to-day advisory work. InsuranceMap has been providing transparency in this area for years. It makes standard insurance policies, coverage, and products comparable. The next step was to combine this carefully curated body of expertise with AI to reduce the burden of administrative tasks, since brokers still spend a great deal of time on them. They read policies, transcribe details, and enter vehicle data. They prepare requests for quotes and compare offers manually. We wanted to develop a technology specifically for this purpose—to process requests for quotes more quickly and generate structured comparisons. This is exactly where InsuranceMap AI comes in. The AI can process information from documents in a structured manner. It can highlight key criteria and clearly compare the incoming quotes. This leaves brokers with more time for advising clients. Especially when insurance companies all adjust their rates at the same time each year—and policyholders consequently want a comprehensive comparison analysis—we can use InsuranceMap AI to respond to these many requests in a timely and detailed manner.
Could you briefly explain how the new quote tool actually works in a broker’s day-to-day routine—from uploading the old policy to the final comparison?
There are essentially two scenarios: If the policyholder wants to compare the terms and conditions of various insurers for the same vehicle, they can upload their existing policy. If a new vehicle is to be insured, the policyholder can also enter the vehicle’s details themselves. In addition, the vehicle registration certificate, the claims history, and other documents relevant to the quote request can be submitted to the AI “for reference.” However, the rest of the process is the same in both cases. The AI analyzes the uploaded documents in a structured manner and creates a draft for quote requests. The broker reviews this draft, adds to or corrects it, and decides which insurers should receive the request. Once approved, InsuranceMap assists with the standardized request for quotes and with consolidating the responses. The quotes received are organized in a way that allows for comparison not only based on price but also on coverage, deductibles, exclusions, and other relevant criteria. The result is a transparent basis for decision-making that can be made available to the client.
They don’t work with a single AI model; instead, they currently combine about fourteen different LLMs. What criteria does the system use to determine which AI to use for which task?
Different AI models have different strengths. For example, one model may be particularly good at extracting structured information from documents, while another may be better at formulating comparisons in a clear and understandable way or at analyzing complex relationships. For us, however, what matters is not the number of models used, but the quality of the results. We select models based on criteria such as accuracy in document analysis, quality when handling multilingual insurance documents, reliability, speed, and cost. A language model alone does not automatically understand the meaning of every nuance of coverage in a Swiss insurance policy. That is precisely why we combine AI with structured domain expertise.
How have you addressed the core issue of data quality in every AI solution? How do you ensure that the comparisons and recommendations are reliable?
For us, data quality is not a secondary concern, but a fundamental requirement. When comparing insurance policies, it is not enough for a system to simply summarize texts in a linguistically sound manner. It must be able to distinguish between benefits, limits, deductibles, exclusions, and conditions with technical accuracy. That is why InsuranceMap operates on multiple levels. First, the comparison is based on a continuously updated database containing structured criteria, standard insurance policy terms, and product information. Second, the AI’s responses are cross-checked against these sources and rules, rather than being generated entirely on its own. Third, the AI highlights any uncertainties or missing information so that the broker can review them specifically.
You mentioned that certain “teething problems” still exist, such as accurately counting damage. How do you deal with such sources of error, and how long does it typically take to correct a pattern?
We build in deliberate checkpoints. If information is unclear or not explicitly stated in a document, the system should not simply guess, but rather flag the issue and trigger a follow-up query. In addition, we record recurring error patterns anonymously and in compliance with data protection regulations, and we continuously improve the extraction logic, validation rules, and user guidance. For clearly defined patterns, corrections can be implemented very quickly. More complex issues, however, require specialized testing with various document types to ensure that an improvement not only resolves an individual case but also functions reliably under real-world conditions.
The AI communicates directly with insurers on behalf of the broker—for example, to request a larger discount. Where do you see the line between automation and the kind of personal negotiation that a broker is actually supposed to provide?
AI is not intended to replace face-to-face negotiations, but rather to help prepare for them more effectively. It can efficiently generate standardized requests, identify missing information, track deadlines, or suggest a desired discount or coverage option in response to a request. This reduces administrative overhead. All communication is managed by the broker and displayed centrally in the chat history. This centralized approach saves the broker a significant amount of time. The broker defines the strategy, reviews the content, approves communications, and takes responsibility for the recommendation.
You’re referring to a traditional AI that is trained exclusively on InsuranceMap’s own database and does not share data externally. How important is this data protection argument for gaining acceptance among brokers and insurers?
Data protection is a key factor in building trust for us. Brokers handle customer and contract data that requires special protection. Therefore, before AI is deployed, it must be clear what data is being processed, for what purpose, where and for how long it is stored, and whether or not it will be used to train external models. Our goal is for InsuranceMap AI not to simply reproduce general information found on the internet, but to operate based on the InsuranceMap library and the documents authorized by the user. Clear technical and organizational safeguards, role-based access controls, traceable processes, and an architecture that complies with data protection regulations are crucial in this regard. To gain acceptance, it is not enough to merely promise data protection. Brokers and insurers must be able to understand how the solution works and where their data remains.
Your pricing structure is tiered, ranging from about 29 francs for a basic AI query to 48 francs per month with upload and tendering features. How did you develop this pricing structure, and which tier is best suited for which customers?
Not every broker needs the same features right from the start. Some are primarily interested in AI-powered expert queries and quick access to coverage details, standard policy provisions, and practical knowledge. Others also want to work with policies and quotes, analyze documents, and handle bidding processes more efficiently. The Basic Level is therefore designed for users who want to use InsuranceMap AI as a professional co-pilot. The more comprehensive tier is designed for brokers who want to integrate document uploads, quote analyses, comparisons, and tender support into their workflow. The financial benefits are quickly apparent, as brokers save hours of administrative work.
You’ve noticed that some brokers instinctively always offer their clients the same insurer. How can a tool like InsuranceMap help change this practice, and does it even require technological solutions—or is it more a matter of a shift in mindset within the industry?
A professional broker should not reflexively offer the same solution every time, but rather consider the client’s needs, the risk, and the specific differences in coverage. Of course, there are situations in which a particular insurer is especially well-suited based on experience, target market, or product quality. However, this should not result in alternatives no longer being evaluated at all. When coverage, limits, exclusions, and premiums are clearly laid out side by side, the real differences become apparent. This makes a recommendation more well-founded, easier to document, and more understandable to the client. We’ve also observed that many customers are no longer satisfied with just one quote. The differences in premiums are enormous. The broker is therefore required to provide the customer with a comprehensive analysis. Otherwise, competing providers could quickly become a problem.
You have a presence on the EcoHub marketplace, where major insurers are also increasingly represented. What role does this presence play in your visibility among brokers?
For us,EcoHub is an additional gateway to the market and a place where insurers, brokers, and technology providers can network. Precisely because the industry is at very different stages of digital transformation, such marketplaces help raise awareness of solutions and facilitate discussions about specific use cases. For InsuranceMap, this presence is above all an opportunity to engage directly with brokers.
How many insurers are currently connected to InsuranceMap via interfaces, and what is the roadmap for further expansion?
We take a pragmatic approach to expanding our interfaces. We already have interfaces with insurers. We have initiated the development of additional integrations with some insurers. We are in discussions with other companies. Other insurers have offered to handle the preparation of quotes on behalf of brokers. This applies to inquiries received through InsuranceMap. This allows us to integrate even insurers without a direct interface into the process. We are expanding the interfaces step by step. Our roadmap is based on actual usage and the specific requirements of our partners.
You view the competitive bidding process in motor vehicle insurance as a starting point. Which other lines of business should be the next to adopt this competitive bidding and comparison approach?
Automobile insurance is an excellent place to start because the processes are highly repetitive, much of the documentation is already organized, and both premiums and differences in coverage are relevant for comparison. At the same time, brokers can immediately see the benefits of a more efficient quotation process. In principle, however, this approach also applies to other lines of personal insurance, such as home contents, personal liability, travel, and legal protection insurance, as well as in the commercial insurance sector. In these areas, InsuranceMap already provides comparison data and specialized information—and is continuing to expand these resources. The order is determined not only by market size but, above all, by data availability, the standardizability of the quotation processes, and the concrete benefits for the advisory process. In the long term, we aim to address areas where a great deal of time is currently lost on documents, data entry, and terms and conditions that are difficult to compare.
Some people refer to such tools as “process optimization,” while others fear a gradual loss of jobs in the industry. How do you assess the long-term impact of InsuranceMap on brokers’ day-to-day work?
I don’t see InsuranceMap primarily as a tool for reducing staff, but rather as a response to growing pressure to improve efficiency and the shortage of skilled workers. In many brokerage firms, repetitive administrative tasks take up valuable time: transferring data from policies, sorting quotes, looking up terms and conditions, formatting comparisons, or following up on inquiries. These tasks will change. Some of them can be automated or significantly sped up. At the same time, the value of tasks that require human judgment is increasing.
Where will InsuranceMap be in five years, and what is the next major step in its development that you can already foresee today?
In five years, InsuranceMap is set to become a central, trustworthy work environment for brokers and advisors to compare insurance policies, ask technical questions, and prepare requests for proposals. It will not be a general-purpose chatbot, but rather a specialized solution built on a continuously updated insurance knowledge base that provides transparent, verifiable results. The next major development step involves integrating individual process steps even more closely—for example, understanding documents, structuring information, identifying coverage gaps, suggesting appropriate comparison criteria, preparing requests for quotes, making responses comparable, and generating a well-founded client document based on this information. Our vision is not fully automated advice without human involvement. Our vision is a broker who can access reliable information within minutes, compare options in a transparent manner, and thereby free up more time for the client. InsuranceMap is designed to provide the professional and technical infrastructure behind the scenes.
Binci Heeb asked the questions.
See also: LINTA by Neutrass: When the Process Thinks
Search:
Sponsors:
