Skip to main content

Reinforcement Learning: AI solutions respond to key decisions

By 3. September 2019August 21st, 2020Dastani Consulting
Reinforcement Learning | KI-Lösungen liefern Antworten auf wichtige Entscheidungen

Behind AlphaZero’s well-known headlines about world-class performance in Go or Chess games is a special form of artificial intelligence: Reinforcement Learning. As with a game, there are also important decisions that follow one another when addressing customers. In connection with the reinforcement learning method, AI technology finds optimal contact strategies for each individual customer.

The well-known AlphaZero algorithm is an example of artificial intelligence that learns independently and improves permanently. This learning process can also be experienced in humans: If a person makes good decisions, he is rewarded – but if he makes mistakes, he learns to improve them over time.

Carrot and stick for algorithms
The idea behind reinforcement learning is relatively simple: the algorithm is allowed to do what it wants to do first – on the basis of a few rules. For games like chess or go, the rules are clearly defined. After a certain amount of time, the system looks at how good the situation is that artificial intelligence has brought itself into. Was the goal reached or not?

Optimizing business results with Artificial Intelligence
The system makes correct and negative decisions – whereby correct decisions receive reinforcement. But also the negative decisions of the algorithms contribute to an optimal successful overall result, because according to a well-known proverb it is said: „From mistakes one learns“. When playing chess, the player also accepts the loss of a piece in order to end up as the winner of the game.

Popular field of application: Addressing customers
The Reinforcement Learning method is well suited for using artificial intelligence to address customers. AI is able to analyze large amounts of data within a few minutes. Reinforcement Learning can put companies in a position to play through various scenarios in order to predict the correct behaviour in the respective situation.

Success of Reinforcement Learning
Our goal is to maximize profit. It is important to know how the customer should be served optimally – by mail, phone, etc., because every contact with the customer is associated with costs. It is therefore important to decide, for example, how many mailings a year should be sent to a customer. The optimal answer would be to address the customer exactly as often as is necessary to achieve the maximum achievable turnover.

Growing wealth of experience from AI
Whether in automation, chatbots or retail – reinforcement learning is the best way to master complex situations on the basis of experience, because artificial intelligence does not forget and its wealth of experience is growing more and more.

Thanks to the method of reinforcement learning, intelligence is really generated and the computer makes much better decisions than we humans are able to. It is not only predicted, but successful strategies are developed.

More precise control of marketing and sales activities
A particularly large lever lies in the control of the sales employees. The costs for a sales representative visit are quite high – between 100 and 400 Euro. The sales costs as a percentage of sales average 15%. It has to be considered when a customer should really be visited in order to use the own resource profitably.

The method can be applied – as shown in our article – to all types of customer contact: in person or by e-mail. For companies this means that the steep learning curve of the algorithms results in a more precise control of marketing and sales activities in order to effectively increase profits in the company.

If you have any further questions, please visit our social media channels (XingLinkedin) or call us at +49 (0)641 984 46 – 0.

 

Dastani Consulting GmbH Im Westpark 8 35435 Wettenberg (bei Gießen) Telefon: + 49 (0)641 984 46 - 0 Telefax: + 49 (0)641 984 46 - 29