Smart communication through artificial intelligence

Self-optimising algorithms

We improve communication strategies for payment reminders through self-optimising algorithms such as reinforcement learning. In this highly dynamic process, the system continuously identifies correlations in the data provided, derives follow-up actions and evaluates the results.

In this way, artificial intelligence independently learns success strategies in fields of application where the variables are both diverse and conflicting. In receivables management, AI helps to harmonise the increase in realisation rates, cost reduction and customer loyalty.

Contextual bandits

The contextual bandits approach is a form of reinforcement learning and is ideally suited for application scenarios with a focus on the user experience. The agent combines exploratory and exploitative approaches for the intelligent management of payment reminders.

Exploratory varies the AI parameters of the communication strategy – for example, the time of dispatch or the channel. The system learns the most effective configuration from the results. This knowledge is then used in an exploitative way to replicate success scenarios and thus make payment communication more efficient.

Deep-Q-Networks

Deep-Q-Networks (DQN) is a novel procedure that combines reinforcement with neural networks. Especially in highly specific application scenarios, DQN enables even higher precision through multiple analysis levels.

Learned correlations in combination with current context data allow the derivation of concrete predictions about the consequences of decisions.

Natural Language Processing (NLP)

We use state-of-the-art technologies such as Stochastic Gradient Descent (SDG) and Google BERT for text recognition. Specially trained intention recognition algorithms enable us to automatically classify customer feedback and thus increase the quality of your customer service.