How AI agents are supposed to save motor insurance

BY ANJA HALL, ORIGINAL PUBLISHED IN GERMAN ON 27 NOVEMBER 2025, VIA VERSICHERUNGSMONITOR

Service provider Solvd wants to process motor claims fully automatically. AI agents handle each step: they communicate with the customer, check coverage, assess the damage, arrange a workshop appointment and even process recoveries – entirely without human assistance. This is intended to help the struggling motor segment become more efficient and reduce costs.
About a year ago, during the quiet days between Christmas and New Year, Andreas Decker had an idea: What if we automated the entire handling of a motor claim? From the initial notification, cost estimation, workshop appointment and payment processing through to potential recoveries – all from a single source.

Decker is Managing Director at Control Expert and also Chief Product Officer at service provider Solvd Group. Solvd belongs to Allianz but operates independently of the insurance group and consists of three companies: Control Expert, GT Motive and Innovation Group. Control Expert specialises in digitalisation of motor claims, GT Motive in damage estimation and cost calculation, and Innovation Group runs a workshop network. In September 2025, Solvd additionally entered a cooperation with HDI Versicherung and took over its service provider SSV Schadenschutzverband.

AI agents work autonomously

Everything was in place to put Decker’s idea into practice. In the first week of January 2025, a development team started work. They are building a platform that uses agentic AI to automate claims processing.

Agentic AI is a relatively new technology that many companies are experimenting with. These AI agents are systems that autonomously pursue goals, plan steps and execute actions. “We no longer define individual processes for the system to run through. Instead, we give the agents the goal of processing the claim in the best possible way,” says Decker. To understand what “best possible” means, the agent receives KPIs and a set of tools. “These can be other agents specialised in certain tasks, but also tools such as invoice checks or cost estimate verification.” The agent decides independently what the next best step is to reach its goal.

Eight sub-agents and a master agent

Solvd uses eight AI agents orchestrated by a so-called master agent. There is an agent for customer communication and one for checking whether the policy covers the loss. Another estimates and evaluates the damage based on descriptions and images. A “context agent” identifies patterns, such as whether a hailstorm is a mass event. Other agents handle workshop appointments and payment processing. One agent filters out particularly complex cases and forwards them to a (human) claims handler.

“In the final stage, a motor insurer can outsource its entire claims handling to us – including recoveries and legal disputes,” says Decker. “The insurer will not need claims back-end systems or staff for claims handling.” The Solvd platform is integrated into the insurer’s systems, billing is done per claim.

Ready in spring 2026

That is the vision. Implementation is ongoing. Around 100 people are currently working on the project, Decker reports, and the team is progressing quickly. Claims intake is already automated, workshop appointments with Innovation Group facilities are also possible. More functions will be activated in the coming weeks. “By the first quarter of 2026 at the latest, we want to be able to handle all claims for a German motor insurer,” says the manager. Solvd will initially focus on bodywork damage. Accidents involving personal injuries and major losses above a certain threshold are excluded for now.

The development did not run entirely smoothly, Decker admits. “These AI agents are powerful models that need to be constrained. We therefore had to define within which boundaries an agent may act and what it may not do.” Task allocation among the agents was also an issue, as well as system performance. At the beginning, it took three seconds for the system to respond. “That was too long. In tests with real callers, people immediately hang up.”

Motor segment under pressure

Decker does not want to talk about current or potential customers. But the project is likely to interest motor insurers because it could solve their most pressing problem: controlling costs. In 2023 and 2024, the segment recorded heavy losses, mainly because workshop costs and spare parts prices exploded.

In 2025, major storms have so far been absent, so motor insurers are likely to return to profitability. But the situation remains tense. Therefore, fast and efficient claims handling is essential. Solvd also sees potential among smaller and mid-sized insurers with outdated claims systems. For them, it may be attractive to rely directly on an external provider instead of spending millions on a new system.

Insurers also do not need to worry that their data might flow to Solvd’s parent company Allianz, the company assures. “We work completely independently for our clients and with strictly separated data environments. Allianz does not have access to the data of other insurers or partner companies.” The platform is set up so that all customer data is technically and organisationally isolated. Data handling is governed by contracts, privacy regulations and independent audits. Internally, compliance and governance processes ensure that neither competitively relevant information nor operational data is exchanged between clients and the owner, explains Solvd.

Not just cost savings, but also time savings

Decker cannot yet quantify how much a motor insurer saves per claim when using Solvd’s platform. In many cases, Solvd does not know the baseline in detail.

But saving time is also a major benefit, he says. The longer a claim remains unresolved, the more likely it is that the customer will hire a lawyer and perhaps request an expert opinion. “That means several hundred euro in costs. You can avoid that if you process the claim quickly and to the customer’s satisfaction.”

Chatbot or human?

But do policyholders even want to talk to a chatbot? The debate comes up repeatedly, Decker says. “Some insurers see personal contact at the moment of loss as the moment of truth in which they fulfil their promise of performance.”
But he believes customers want something else. Solvd conducted a survey asking policyholders what matters to them in a claim. 82 percent of callers want their problem solved immediately, says Decker. “Whether an agent or a human does it – they do not care at all.”