Automation was a big issue long before digitalization, but it has now picked up speed again. All companies want to automate. The demand is enormous. It is forgotten that the next step, autonomisation, is just around the corner and will become a game-changer. Straight and above all for the software industry.
Autonomization vs. Automation
Many people cannot distinguish automation from autonomy. But it’s simple: an automated system can execute predefined processes without human intervention. If a deviation from the standard process or a completely new requirement occurs during processing, the system stops and requires the human operator to clarify the situation.
Autonomous systems, on the other hand, are designed to carry out different types of work and processes independently, i.e. autonomously, in one department. The system does not need humans.
Autonomous systems are “game changers” because they radically reduce process costs. At Parashift, we have developed an engine that already delivers 100% correct results for data extraction and posting of accounting documents. Compared to automated systems, e.g. for invoice processing, which have a sophisticated exception management system, companies using this service have the advantage that no employees are needed at all for the correction and validation of the data. This makes the process cost curve bend in the right direction.
Building autonomous systems is a marathon.
While the advantages of autonomous systems are obvious, building autonomous systems is extremely complex. This is mainly due to the fact that sooner or later the machine has to find the best possible solution for every situation (according to all usages and rules of the department). By definition, it is impossible to cope with this complexity with a set of rules. It will never work. This is probably the reason why in the accounting industry the idea that bookkeeping cannot be autonomous is persistent.
We solve this problem by constantly developing our own machine learning platform, which learns from a gigantic amount of processing data. It is not sufficient per se, for example, to have only accounting records, but the interaction data of qualified, human processing are of greater value. We are currently training the platform specifically for case accounting, but the process and technology behind it could basically be used for almost any form of autonomous handling of qualified work.
In its infancy
Since the task is complicated and usually enormously complex to solve, autonomous systems are extremely rare. And I am of the opinion that systems that do not learn independently and/or cannot apply this knowledge in context to new cases are not real autonomous systems. They are then rather perfect and broadly automated mechanisms. It is only a question of time before a new requirement arises in such a scenario, which the system cannot process. Translated into business, this means that it will no longer be possible to lower the total processing costs at one point.
Autonomous systems are like children
In the further, I experience again and again, how even very innovative people from IT are arrested in a conventional software understanding. Autonomous systems are not software projects, which you start, develop and which can deliver results XY.
Rather, these systems are like a child that you first give birth to and that can learn relatively little but at the beginning. You feed it, start learning more and more complicated things, encourage and challenge it, set limits and spur it on. This is a completely different approach to software development. By definition, such a project has no end. We try to establish this philosophy more and more in our team.
From tool to a service partner
But autonomous systems have what it takes to turn the software industry and the world of work upside down. Autonomous systems change the role of software.
While software today is a tool to make work faster, easier, cheaper and better, autonomous software takes over the work completely. To stay with accounting: While today accounting systems are used to let people do accounting, future autonomous systems will make accounting independent. We are much closer to this point than people generally think. We are working at full speed on it.
When software controls real things
Logically, autonomous systems have a particularly large impact when they begin to control objects of our daily lives. The most discussed example in this respect at the moment is certainly the autonomous vehicle. I think the combination of autonomous software and physical things/machines in a cadence will create undreamt-of new possibilities that we can’t even imagine.
At the very beginning
We are at the very beginning and I don’t know many people who know or deal with general concepts for the autonomization of software systems. This probably has to do with the fact that the sum of challenges and the level of complexity are overwhelming at first. However, the more one deals with it, the more one realizes that for many of these challenges there are relatively simple approaches that work well. What was extremely difficult for us as a start-up was to build a business model around the ongoing development of the platform that would help recoup and ultimately offset the development costs.
After a period of extreme uncertainty, the search for possibilities, we now believe we have found an economic model that can be generally used to develop complex autonomous systems. This does not mean that it will be easier to achieve the technological goal. But as we all know, there are no shortcuts to places that are worth going to.