Many organizations are run with ERP systems. The beauty of these systems is the integration of numerous business relevant functions.

Stuctured Data

When such systems are being implemented, processes of the organization will be mapped onto the processes supported by ERP system and operational data needs to be harmonized to the underlying data structures of such ERP system. When an organization goes live with an ERP system, all reports, lists, key metrics, data sheets or analysis will have to come out of that system.

The downside of such systems lies in the complexity of its data structures. All too often, especially in international organizations, the data needs and reporting requirements cannot be met by the standard reports that the ERP system offers. Customized data extractions and data preparations, typically using Microsoft Excel, diminish the value of the ERP system implementation.

Users are left to handle abstract and complex data structures with pivot tables, lookups and various formulas. Large spreadsheets without any sort of integration or data consistency mechanisms are the inevidable outcome. Cost calculations, headcount statistics, sales projections, bill-of-materials, procurement lists or capacity planning applications are typical examples for data that is maintained in an ERP system and then used and ammended in distributed Excel spreadsheets.

In addition, user roles and access rights will be compromised when increasing request for data extraction and downloads become routine business. Data ownership principles often fall short of the rigid rules that preserve data integrity. Loose access and maintenance rights open doors to erroneous data, lost data and duplicate data sets. After a while, even the data in the ERP system lacks consistency and integrity.

Non-structured Data

Apart from ERP Data and the many Excel spreadsheets to support it, there are many more data sources in an enterprise. Think of all the data that is being generated on the production floor, think of all the communication that is done on the shop floor. Or think of all the data that can be obtained by simply looking at human behaviour. When you think of dates, names, product details, transaction information, and so forth, you know that you have structured data in mind. At the same time, unstructured data has many apperances like text files, PDF documents, social media posts, comments, images, audio/video files, and emails, to name a few. Structured data is highly specific and is stored in a predefined format, where unstructured data is a conglomeration of many varied types of data that are stored in their native formats. Some data however, is not yet captured and stored at all.

Imagine that you would video-record yourself, when you communicate with someone. You would get the words you say, the way you say it. You would recognize body language and facial mimic. You would notice spatial details like approach, distance, space used etc. These are details nobody (except for maybe the security services branch) is capturing. However, if you believe that one day we will see machines that want to communicate with us, think for a moment about what data these machines must have been fed, in order to learn about communication. Exactly. The only way machines can learn to interact with humans is by studying humans.

There are already endless surveillance tapes on this planet, showing people in all kinds of settings: at airports, in restaurants, in public spaces, in shops etc. The data is available as a video file, hence it is unstructured data. But so is a basket of lemons or a chessboard with some figures left. What makes machine learning so interesting is that these machines learn even from unstructured data. At present, this kind of learning is to a large part based on pattern recognition, outlier detection, extrapolation and prediction. Soon we will see machines enter the arena of judgement and decision making, thereby producing new data, such as failure rates, errors and loss. But unlike the machines mistake-shy human counterparts, machines see this only as more data to improve what they are doing. And they are doing it fast. Given the exponential learning curve, we can expect machines to one day bypass our capacity when it comes to judgment and decision making - most prominent example being the already in place mandate for some doctors at some hospitals to consult the AI for analysis before proceeding.

Data for machines to learn from

On a larger scale every interaction between humans will be an example for machines to learn from. Having said this, it becomes clear that whatever we do, be it nice or naugthy, will become "training data" for any learning machine. Worst case scenario: a machine learns that humans don´t stick to their own rules, so the machine concludes that this is ok. But hopefull we are still far from this scenario.

In the meantime however, we companies should prepare for another highly likely scenario, in  which competitiveness will be driven by the company´s AI´s capabilties. Car manufactures that use an AI to find the safest, most ecological and also highly attractive car design, will likely have an edge over their competition. Pharmaceutical companies that use AIs to analyse new viruses and swiftly design new drugs and vaccines using their AI will have a hefty time-to-market advantage. Cities that feed all their surveillance data to AIs in order to get insights and predictions about crimes and desaster will be able to reduce fear amoungs its inhabitants and thereby generate massive attraction for rich people.

While the Googles, IBMs and Alibabas of this world are already in the race to build the first Superintelligence, the mom-and-pap stores and even the mid-sized companies will have to wait what they come up with. For the bigger companies however, the race is on. With hundreds and thousands of employees the data that these people produce is enourmous - and again, we are not just talking about structured data in the ERP systems. Big corporations have a vast opportunity to become part of a big data race, that will generate gazillions of gigabytes for machines to learn from And learn they will.

Be prepared

We can help you with modeling the data that is needed in order to fulfill daily operations requirements and in streamlining the processes that allow to work with data out of your ERP system, without compromising it. We can also assist in migrating data from Excel spreadsheets into SAP, in extracting data from heterogenous ERP systems as well as in creating large Excel spreadsheet applications that sustain data integrity. We can also assist you in or manage your business intelligence project, building ETL paths and OLAP applications for fast analysis. But none of that is new.

We like to help you make the first steps into a new age, where the collaboration with machines is the predominant scenario, where your data is the key to the capabilties of these machines (because they learn from it) and where your competitive advantage is nothing short but dependend on a solid MACHINE LEARNING DATA STRATEGY.

There will be algorithms available shortly, that will enable you to test your data against various methods. As a result you will receive predictions and other insights - conclusions from the analysis of your data. We can help you get access to such approaches and in preparing your data for using them.

In the not so far future however, your company will be able to understand and develop its own algorithms. Perhaps your company is already running an AI department and several projects. According to many research results the majority of such teams are only getting their feets wet. What is lacking is a sound approach to establish an AI BUSINESS CASE. We would be happy to assist you with this.