What is ‘true automation’?

/ by
Reading Time: 3 minutes

Automation might seem cutting edge, but it actually predates computing.

The word traces its roots back to 762 BC when Homer’s Iliad famously recounted the tale of Hephaestus, who relied upon the help of self-operating machines called ‘Automatons’ to manufacture weaponry for the gods of Mount Olympus.

Fast forward two and a half millennia, and automation became a widespread phenomenon during the Industrial Revolution, with the invention of the automatic loom. And today, the press is saturated with stories of how automation is going to refashion the global economy.

Replication and non-replication

The first approach to achieving automation seeks to directly replicate human activity, for example, by having a robotic arm move pieces on a chessboard, or using Robotic Process Automation (RPA) to simulate a human at a keyboard by literally typing on the keys. This might seem crude, but that’s because the task being automated requires an interface built for human beings. For example, RPA is used to automate interaction with a computer application which can only be navigated via the buttons and text fields presented to a human user.

The second, most common (and sensible) approach to automation, is to allow the underlying technology to do what it does best, without pretending to be human. This is how most IT works today. Human problems to be solved by computers are stated in terms of use-cases and solved in computational ways. Computer Science has progressed tremendously in terms of how computers can be programmed to solve real-world problems and how the mass of computational logic in any IT system may best be organised and maintained by human overseers.

In mainstream applications, computers and their programs don’t need to replicate the process of human thought in order to deliver solutions. When an IT systems provider thinks about how to automate and computerise a human process, they consider how computers naturally operate, not how humans naturally think. And if the requirement is to automate another computer application, the target application presents a computer-to-computer interface rather than a human-to-computer interface, in order to allow this non-human interaction to take place.

The onus is on us as technologists to translate human problems into ones that can be solved by computers. However, there’s a whole class of problems for which this is simply beyond our current ability to analyse and synthesise.

Artificial Intelligence

With the recent increase in computational power, interest is intensifying in automating non-deterministic processes by attempting to replicate the functioning of the human brain. This is where the third approach to automation comes in.

Neural-network systems – generally referred to as AI – are making great headway in areas such as facial recognition, language translation and recognition of patterns of human behaviour, such as fraud detection.

Although exciting to IT professionals, it’s difficult to think of this approach as anything other than a magical sledgehammer. AI systems are fed with masses of learning information which they assimilate into some internal database in order to answer future questions that are then put to them. Answers might be right or wrong and technicians have no idea how they’re reasoning, so even if AIs get things right today, they may get them wrong tomorrow. Just like people, really.

Automation and document interpretation

Success in automation means picking the right approach to the right problem and applying technology so that human involvement is removed altogether.

Without denigrating the work taking place within RPA and AI, no one would ever dream of using these techniques to automate a process that isn’t hamstrung by a human-only interface and whose underlying logic can easily be analysed and programmed into software.

Truth be told, automation exists by degrees. When people are assisted in their work by machines we see improvements in processing speed and efficacy, but the biggest benefits occur when humans are removed from the process altogether. Only then can a service be delivered at the speed of the underlying machinery, which in this day and age, means the speed of a computer. This we call ‘true automation’.

Document interpretation presents us with fertile ground for true automation, since a document’s text and properties are generally available via a computer interface and the logic defining the meaning of that data is easily understood and expressed. Many technology providers in the electronic invoicing sector have latched onto strong client demand for automation, but few can actually deliver it in its purest form. In fact, most companies that claim to automate sales order processing and invoicing simply augment data-entry, failing to help clients truly achieve automation and associated cost savings.

CloudTrade is different. Rather than simply extract data as a lot of systems do, our engine uses patented technology to extract and interpret information carried within documents. This allows a human rules writer with minimal training to quickly and easily automate the way we understand document data, allowing documents to fly through our engine at ‘computing’ speed without human intervention.

This is true automation, delivering meaningful costs savings and operational efficiencies for clients large and small.


Here’s how CloudTrade automate your PDF documents

CloudTrade Video - Automate your back office