March 6, 2024

At Allianz Trade, we have always maintained that our focus is you – our valued customer. In this rapidly evolving landscape, we embrace cutting-edge technologies like artificial intelligence (AI), but only to elevate the experiences of our customers.

What sets us apart is our unique approach: AI serves as a supportive decision-making tool for our in-house experts, including analysts and underwriters. It strengthens the information gathered from customers, reinforcing our commitment to personalized service. However, a conversation with an Allianz Trade expert continues to be the most effective way to conduct business.

Héléna Bergez
Global Head of Credit Assessment
Allianz Trade

Héléna Bergez, Allianz Trade’s Global Head of Credit Assessment, explains how we are using technology to enhance our customer-centric approach, helping us to focus on buyers with higher uncertainty.
For example, we’ve introduced an FAQ chatbot, which addresses straightforward buyer questions faster and more efficiently. Additionally, our automated Yseop tool can explain our credit decisions in great detail. We use these solutions alongside machine learning models to optimize the productivity of our underwriters, giving them more time to concentrate on human interactions and assess riskier buyers.

According to Héléna, giving our customers the confidence to trade with such buyers provides them with a competitive edge.


We have also made great efforts to integrate structured and unstructured data into our processes, allowing us to take a substantial leap forward in predicting company defaults. For example: 

  • Our internal machine learning-powered grading tool enhances our grading system by automatically assessing buyer credit risk on a scale from one to 10 based on in-house data. Known as Shamrock, it excels in evaluating companies despite limited data, streamlining processes across various countries, including the UK, France and the US.
  • Machine learning helps to detect fraudulent buying patterns. Our algorithm identifies buyers with the highest fraud probability, alerting analysts so they can conduct a thorough review and decide the best course of action. 
  • We are also automating manual workloads for underwriters who handle large volumes of credit limit requests with varying complexity. Our automated solution, driven by statistical learning and business rules, automates obvious decisions with high precision and minimal impact on expected loss.

Héléna underscores the power of machine learning-based tools in improving grading performance, increasing acceptance rates, accelerating response times, and enhancing coverage with more accurate grades. 

In short, we are using machine learning technologies to significantly enhance efficiency and reduce the overall cost of doing business, allowing our analysts and underwriters to focus on areas where their expertise has the greatest impact. 

For example, we have trained our machine learning-powered fraud prevention tool to automatically recognize various fraud risks. These include identity theft, buyer swindles, fake transactions and false claims.

Common signs of potential fraud include: 

  • Unusually high-intensity limit requests from specific customers
  • Buyers registered in high-risk geographic locations 
  • Several annual financial statements received at the same time 

It’s important to note that our machine learning fraud detection does not automatically adjust buyer gradings when suspicious activity is detected. Instead, our expert analysts receive alerts and use their knowledge to investigate these warning signs thoroughly. Action is taken only when the evidence is compelling, ensuring a proactive and precise response to potential fraud risks. 

As customers and their buyers may question our grading decisions, demanding detailed explanations for each decision could prevent our analysts' ability to focus on other critical tasks. That is why we have developed an innovative solution that plays a crucial role in ensuring a seamless customer experience.

This sophisticated tool steps in by automatically analyzing customer features and sector data, extracting key financial information, comparing it against optimal financial parameters, and generating a comprehensive narrative that explains our grading decisions.

Héléna explains that the beauty of this tool lies in providing explanations, for instance linking an attributed grade with factors such as significant turnover decrease or being a subsidiary of a financially vulnerable parent organization. The tool allows our analysts and underwriters to concentrate on more value-added customer interactions.

While this tool and our FAQ chatbot serve as the initial customer interaction touchpoints, they are complemented by our commitment to personalized service. When clients or buyers seek further clarification, they are connected to one of our in-house subject matter experts, ensuring a detailed response.

Operating within a highly regulated environment, Allianz Trade diligently monitors all machine learning decision support models in compliance with fast-evolving regulatory requirements.

However, our approach to data usage goes beyond compliance – it’s a conscious choice rooted in a risk-based strategy. Notably, we refrain from using confidential data or information from unofficial sources in our machine learning models, private individuals are excluded from our models, and confidential data is never externally displayed as model outcomes.

In light of this commitment to clients and their buyers, we have developed a new generation of machine learning models with reinforced transparency and explainability. One such solution has just been launched in the UK market, with expansion planned for nine additional countries within the next 24 months.

The results so far are incredibly encouraging,” says Héléna. For example, our latest machine learning solution utilizes fewer data points but is more accurate thanks to its increasingly advanced algorithms.

Our new machine learning generation ensures that buyer-grading decisions are not only accurate but also easier to explain to both customers and regulators. In a nutshell, it ensures the total predictability and explainability of each model variable.

This aligns with our commitment to responsible and transparent use of technology. 

Our machine learning use cases share a common goal: enhancing the capabilities of Allianz Trade credit analysts and underwriters. By strategically incorporating powerful machine learning solutions, we aim to amplify their expertise, allowing them to intensify their analysis of crucial buyer cohorts and sensitive cases while dedicating more time to interacting with our clients.

Embracing advanced technology is key, but Héléna reiterates that the core of our customer value proposition lies in nurturing client relationships and providing direct access to human expertise.

Our in-house experts concentrate on client relations and the handling of sensitive cases, ensuring a personalized touch in our customer interactions.