{"id":2743,"date":"2024-06-27T09:04:44","date_gmt":"2024-06-27T07:04:44","guid":{"rendered":"https:\/\/www.d-velop.com\/blog\/?p=2743"},"modified":"2025-09-22T08:43:50","modified_gmt":"2025-09-22T06:43:50","slug":"mail-distribution","status":"publish","type":"post","link":"https:\/\/www.d-velop.com\/blog\/process-automation\/mail-distribution\/","title":{"rendered":"Please only distribute mail intelligently in the future – the district town of Bergheim shows why!"},"content":{"rendered":"\n

If SpaceX wants to colonise Mars in 2025, then we should be able to handle all administrative processes electronically by then, right? How can it be that the planning for colonising Mars is already in full swing, but here on Earth, we are still distributing huge mountains of paper mail manually?<\/p>\n\n\n\n

The district town of Bergheim\u202fasked itself (almost) exactly this question\u202ftogether with\u202fcodia Software GmbH\u202fsome time ago. By introducing\u202fe-files\u202fin administration, the right answer can be found quickly:\u202fintelligent incoming mail distribution.<\/p>\n\n\n\n

In order to understand why and when mail distribution can be described as intelligent and what this actually means for administrative practice in the future, a brief excursus on the topic of\u202fartificial intelligence<\/a>\u202fis necessary.<\/p>\n\n\n\n

<\/span>When exactly do we talk about \u201cartificial intelligence\u201d?<\/span><\/h2>\n\n\n\n

When you talk about intelligence, you most likely picture a human brain. This consists of a large number of nerve cells called neurons. These are connected to their neighbouring cells via many paths. Electrical impulses are constantly sent from one cell to another via these connections, enabling people to learn, reason, and think abstractly.<\/p>\n\n\n\n

Basically, in so-called \u201cartificial intelligence\u201d \u2013 also called \u201cAI\u201d \u2013 the neurons are represented by algorithms and computer codes.<\/p>\n\n\n\n

A general distinction is made between strong and weak artificial intelligence. The latter has been an integral part of our everyday lives for many years. In 1997, the reigning world chess champion Garry Kasparov had to admit defeat against a computer, that Siri can tell us what the weather will be like tomorrow and whether we need an umbrella or that annoying emails in the spam folder land.<\/p>\n\n\n\n

All of these examples have one thing in common. They \u201conly\u201d specialise in the intelligent solution of one or a few tasks. Weak artificial intelligences are not yet able to work outside the context for which they were programmed.<\/p>\n\n\n\n

There is still a long way to go before a \u201cstrong artificial intelligence\u201d that can be compared to the intellectual capabilities of humans emerges. <\/p>\n\n\n\n

<\/span>How can the machine manage to beat the human?<\/span><\/h2>\n\n\n\n

First of all, the system must be taught how to learn on its own. This requires many artificial neurons. In the so-called learning phase, the system is fed millions of data elements that needs to be analysed and classified. During this time, it constantly receives feedback as to whether the processes were carried out correctly or not.<\/p>\n\n\n\n

Based on this feedback, the system changes the paths between the artificial neurons. Quite similar to humans. The connections that led to the \u201cright\u201d result are strengthened. The \u201cwrong\u201d connections, on the other hand, become weaker.<\/p>\n\n\n\n

After a large number of tests, the system should develop into an intelligent, neural network that can further optimise itself independently. It is now able to correctly classify completely new situations based on the connections established. This process is known as \u201cmachine learning\u201d.<\/p>\n\n\n\n

This makes it possible for a machine to far exceed the capabilities of a human in this very specific area. But only in this area.<\/p>\n\n\n\n

<\/span>What does artificial intelligence have to do with the distribution of incoming mail in the district town of Bergheim?<\/span><\/h2>\n\n\n\n

The district town of Bergheim would like to fully and consistently automate the receipt of its mail distribution. Artificial intelligence is already helping to accelerate and optimise this process.<\/p>\n\n\n\n

What is currently happening in the city of Bergheim?<\/h3>\n\n\n\n

The \u201cmountains of mail\u201d arriving at the administration of the district town of Bergheim are scanned and imported as PDF files into a simple and fast document processing system. A text recognition system (OCR) extracts all the words from these documents. The document is then automatically assigned based on these words, archived and forwarded to the correct mailbox.<\/p>\n\n\n\n

At this point, the first important prerequisites for the path to a completely intelligent inbox are already being created.<\/p>\n\n\n\n

The positions of the individual words in relation to each other or patterns are automatically recognised by the system and assigned to the departments according to predefined rules. This makes it possible for the recipient to work with the document directly via their mailbox in the ECM system, whilst assigning invoices to appropriate workflows.<\/p>\n\n\n\n

The challenge is that these documents are not just standardised forms, but freely formulated and unstructured documents.<\/p>\n\n\n\n

\"\"\n\t\t\t\n\t\t\t\t\n\t\t\t<\/svg>\n\t\t<\/button><\/figure>\n\n\n\n

<\/span>How do you determine which mailbox the document should go to?<\/span><\/h2>\n\n\n\n

Automatic assignment of incoming documents is possible because \u201crules\u201d have been created beforehand, these representing the connections between the individual mailboxes.<\/p>\n\n\n\n

Each department can be assigned concise keywords and linked to rules. Such as: \u201cIf this word appears in a text, then the document is intended for department X\u201d.<\/p>\n\n\n\n

Text recognition enables filtering on selected words. Any document in which this word appears and is recognised, can be automatically forwarded to the responsible digital mailbox<\/a> through the mail distribution receipt.<\/p>\n\n\n\n

This logic can be extended as desired. This makes it possible to link \u201cand\/or rules\u201d with each other: \u201cIf this AND that word occurs, then the document belongs to department Y\u201d or \u201cIf this OR that word occurs, then the document goes to location Z\u201d.<\/p>\n\n\n\n

In this way, the correct individual rules can gradually be created and used for mail distribution in the district town of Bergheim. Important: The rules are not specified, but can be determined by the respective user depending on their needs.<\/p>\n\n\n\n

<\/span>Can the system learn from mistakes?<\/span><\/h2>\n\n\n\n

This is exactly where artificial intelligence will be used in the future. The goal is to create a self-learning system in the near future through \u201cmachine learning\u201d.<\/p>\n\n\n\n

It should be able to determine if an executed process did not lead to the correct target mailbox. The intelligent system itself must then be able to correct the connection, so that the document reaches its destination.<\/p>\n\n\n\n

At the moment, this error is still being recognised and adapted by people. But it will soon be possible to have these processes carried out automatically and, above all, by the intelligent system itself.<\/p>\n\n\n\n

<\/span>The benefits of a digital mailroom<\/span><\/h2>\n\n\n\n