30.000 options, 1 goal.

When customers don't pay, we use a combination of artificial intelligence (AI), data science and behavioural science.

Our self-learning algorithm typologises people based on data and actions to execute the most appropriate strategy.

In respectful interactions, our AI sensitises the person to pay and finds a customer-focused solution.

Artificial Intelligence.

We use the most advanced AI technologies to create a highly personalised customer experience in debt collection.

Generative AI Reinforcement Learning Supervised Learning Natural Language Processing

Generative AI is based on Large Language Models (LLMs) and is particularly suited for automating inquiry processing in inbound communication within debt collection. It can independently categorise requests like instalment plans, payment deferrals, or disputes, as well as selecting and sending the tailored response. It can decide whether a response should be handled by human staff or the generative AI itself. This significantly accelerates the communication process.

Reinforcement Learning (RL) is suitable for choosing the best strategy. If a strategy is successful, the model is rewarded, continues to learn and optimises it. RL enables a personalised customer approach in debt collection by taking individual preferences and behaviour patterns into account. Deep Q Learning is a complex form of reinforcement learning based on neural networks. It can be used to derive optimal actions in the long term and determine, for example, the best time or the best communication channel for a payment message.

Supervised learning is used to predict concrete events and often also their probabilities. In the process, labelled data sets are used to train algorithms. Supervised learning is a form of machine learning and can be used in receivables management, for example in scoring. Thanks to supervised learning, the algorithm can estimate how likely it is that a person will pay the outstanding debt.

Natural Language Processing (NLP) enables large amounts of text to be processed automatically and relevant information to be extracted. This can help to process customer enquiries efficiently and in a timely manner. Large Language Models (LLM) such as Llama 2 can analyse emails or text messages, for example, and recognise whether a payment agreement is desired, there are queries or a complaint needs to be prioritised. In concrete terms, the LLM understands, for example, that a consumer wants to conclude an instalment plan in 3 instalments.

Data Science.

Our algorithm processes individual characteristics of your defaulting customers, which are obtained from data analyses before and during the interaction in the dunning process. Based on these characteristics, our AI identifies certain behaviour patterns and specifies the typology.

First-hand customer data

Your customer data gives us insights into consumer behaviour as well as indications of systematic behaviour patterns of your customers and determines the communication channels.

Third-party data

Data from credit agencies, credit institutions or market data help us to get a better picture of the solvency of your defaulting customers. This enables us to offer targeted, individual solutions.

Digital data

Data obtained from social media or in the context of email identification also provides information about the socio-economic status of your customers and also determines the communication in the dunning process.

Behavioural data

The accessibility, the type of response or the reaction speed of your customers provide information about their willingness to pay and show us whether the measures in the dunning process are effective.

Behavioural science.

Based on findings from behavioural research, we determine the optimal contact with your customers. During the interaction, our algorithm adapts the communication individually, depending on the customer's reaction, until we reach an agreement with your customers.

The advantages.

Book a demo

PAIR Finance

Technology

Markets

Company