Process automation must not be implemented in isolated solutions, instead the whole end-to-end process view must be considered.

Companies use Intelligent Automation to have computers carry out transactional tasks. Let's take a look at the most common types of Automation:
 

• Robotic Process Automation (RPA) operates software via the user interface, like a human user, only faster and more persistently
• Optical Character Recognition (OCR) recognizes text content from scans and makes the data usable for machine processing
• Natural Language Processing (NLP) such as chatbots and voice recognition enable the users to interact directly with the system
• Artificial Intelligence (AI) helps to structure data meaningful and evaluate it efficiently
  

However, during the implementation there are four typical observed mistakes:

Often, individual company departments do not coordinate their digitization strategies, resulting in different uncoordinated systems. Purchasing, Sales, Finance, Logistics - each department has its own isolated solution. The potential of automation is not nearly exhausted and the return on investment falls short of expectations. An overarching strategy is missing.

Automation is comparable to an orchestra. The individual instruments can also play melodies, but only when you put everything together and have someone to control the operation, it becomes a round, sonorous piece.

How to do it better?

Process Automation must be viewed from an end-to-end perspective. An example for end-to-end automation is the following: OCR extracts information from a scanned invoice. The data flows into a database where Artificial Intelligence, e.g. IBM Watson, recognizes the content of the data and assigns it accordingly. Watson then forwards the data to a RPA program to post the transaction. It is not required to introduce the whole process at once, it can be also implemented gradually. For example, RPA can be implemented first, to carry out standard bookings or order initiations to realize efficiently cost savings. In a second step, the Artificial Intelligence is set-up and trained in the background.

It is important to keep a transparent overview. At what point in the process is a specific booking process required? Anyone who interlinks different systems but does not make them transparent will have problems to identify process errors. It is vital to keep the overview to identify deviating customers, orders, or bots that cause later manual rework.

How to do it better?

Intelligent Automation reduces transactional effort, but increases the monitoring and, if necessary, correction effort. However, this can also be done by Artificial Intelligence. Well thought-out automation includes automatic controls. When properly calibrated, it detects deviations and can independently correct faulty versions. It can anticipate where and when which resources will be needed and controls them accordingly.

If individual applications that are used by several bots must be maintained (even unexpectedly), the bots should of course not produce any errors. Instead, companies must create a central point that prevents thousands of processes from going wrong at once. Defined waiting points require bots to pause and process the accumulated processes in a structured manner after the maintenance work.

The right employees are crucial for Automation. Bot providers often promise the simplicity and implementation ease of their products. However, it is only one side of the coin. Without personal support, appropriate training and ideally experts in the background, introduction and operation are usually not possible.

How to do it better?

An end-to-end automation strategy also includes determining the personnel requirements: What competencies do I need to guarantee technical and professional operation? Object-oriented programming can also be conducted by departments, but the knowledge must be built-up first. Each company also has its own IT requirements that must be met. Of course, it is also possible to outsource the services, but you won’t succeed without dedicated specialists.

There is a good reason for the four-eyes principle, which is also often required by law, to prevent that someone will harm the organization and enrich themselves. We observe often that customers give developers more authorizations than the internal guidelines would have allowed. In addition, the programmers occasionally did not use a development environment, but developed directly in the production system.

The superiors only became aware of the dangers after the process risks have been explained to them. A common example: The same programmer has the option to create a creditor, record and execute a posting. Supervision? Non-existent.

How to do it better?

It is not recommended to trust the programmer only. Companies should pay close attention to given access rights during bot development. It is advisable to rely on so-called “pairing” in critical areas. One programmer is always controlled by a second one. After a few hours, the person previously controlling takes over and is now monitored by a third programmer. This excludes hidden commands and at the same time ensures that the code is as efficient as possible.

Suspicious process anomalies can also be monitored by Artificial Intelligence. Does a bot need significantly more capacity than usual? Does the bot access other applications? Is the amount of processed data unusually large? Is there an external data drain?

In any case, companies that automate processes should create a clear role and reporting concept to ensure stringent compliance.

If you pay attention to these four recommendations, you can implement Intelligent Automation efficiently and benefit from higher returns.