PAT – The first AI in the Print Media Industry

PAT – The first AI in the Print Media Industry

Artificial Intelligence (AI) is still something that sounds very futuristic for the most people. But it is already there! Tools like Siri or Alexa listen to us in every moment and provide answers to all of our questions or even help us in turning lights or music on and off. Navigation systems show us the best and fastest way to our target destination while in parallel continuously monitoring the traffic and suggesting alternatives in case situation changes. Even if we sometimes don’t recognize it, it is already all around us. And it helps us a lot!

The precondition to make use of an Artificial Intelligence is always the availability of data. AI needs data to understand and learn. Without data it will just be a simple feature. But if you feed it with data, it will learn. And it learns fast!

As Artificial Intelligence is already used in many industries and helps to improve processes, performances, and many other things, it is also a very valuable technology for the print media industry.

With several thousand sheetfed presses connected to the Heidelberg Cloud we are receiving an unbelievable amount of data every day! We are talking about the data of hundreds of thousands print jobs a day. And every print job provides us a lot of information about the performance and the technical condition of the press. These huge amounts of data would simply not be useful for us without the help of Artificial Intelligence.

With PAT, the Performance Advisor Technology, we created an Artificial Intelligence that is able to manage all this data. While looking at the performance of the press, PAT automatically recognizes significant changes and provides explanations and recommendations to help our customers to improve the overall performance.


To do so, PAT is mainly working in 5 steps.

Step 1: Observe the pattern

PAT is not only monitoring the upcoming data but is also looking at past performance indicators or, with the help of Performance Benchmarking, he can even compare the performance of the customer with anonymized data of other presses printing the same jobs. With that he can define the average performance of a certain press based on their production profile. He then uses this information to define case specific thresholds which are used as trigger for upcoming notifications. That means PAT is able to recognize significant changes in relation to the specific production profile of the customer.

Step 2: Expert rules

After PAT has identified changes in performance data, he comes up with specific recommendations which help the customer to improve the performance again. To do so, PAT provides recommendations to the customer within the Heidelberg Assistant. These recommendations are defined by our Heidelberg Experts (human input) and will help the customer to understand and solve upcoming issues.

Step 3: Recommendation Applied

After the customer implemented the recommended action, he simply closes the action and informs PAT automatically about the exact time the recommendation was applied.

Step 4: Observe Impact

After the recommended action was applied, PAT observes the impact. That means, he is monitoring the change of the specific data and understands if the recommended action helped to improve the performance or not.

Step 5: Continously improving the rules

Here comes the intelligent part. By observing the impact, PAT understands which recommended actions help significantly to improve the performance. With that he modifies the priority. In case an action helped to improve the performance, PAT learns that this action is a good action and he increases the priority and recommends it more often. In case a recommended action didn’t help to improve the performance that much, PAT will also understand that and decrease the priority. That means that PAT is learning from case to case.

As this information is also forwarded to our experts, PAT helps us to understand which actions really help to improve the performance or solve an issue. This enables us to exchange or modify the existing recommended actions.

This Process will be repeated with every upcoming case, which is why a high amount of data is important, as it gives PAT the possibility to learn. With every new case, PAT learns a little bit more and understands which recommendations really help our customers to solve issues or increase performance. In a long term that also means that all recommendations provided by PAT are useful, as all others have already been sorted out.

Of course, at the moment we are still at the beginning, but PAT is already getting better each and every day and will soon be a tool no one want to miss again!

Related articles

By continuing to use the site, you agree to the use of cookies. more information

The cookie settings on this website are set to "allow cookies" to give you the best browsing experience possible. If you continue to use this website without changing your cookie settings or you click "Accept" below then you are consenting to this.