By Marco van der Hoeven
Many organizations collect huge amounts of data, but without a clear underlying strategy. But just like oil, which data is often compared with, raw data is worthless in itself. Something still has to be done before innovation is possible with the information that comes from data.
“Many organizations would like to work data-driven, but have enormous challenges in finding the necessary data,” says Bart Buschmann, Commercial Manager at eMagiz . “Strangely enough, they often already have it inhouse. Recent research by Gartner has shown that approximately 70% of the data collected by companies is actually never used again. They call it Dark Data.”
A benchmark that eMagiz conducted itself among large Dutch enterprises confirmed this picture. “It turned out that at least 50% of the data is ‘untraceable’. Very often they probably don’t even know they have that data. In that respect, I sometimes compare data with oil. When extracting oil, a lot has to happen before you have the crude oil on the surface. You have to find the oil fields first, then drill through very thick rock layers and build pipelines.”
Value
“And when you finally get the oil to the surface, you have an intrinsically worthless product; you have to do something with it before it produces value. And that’s exactly how it works with data. In the past years storage has become so cheap that everyone tends to store and keep everything. As a result, we have immense amounts of raw data, but a lot still needs to be done before this data yields information that is of value. And the danger of this situation is that even newly generated data is immediately lost, because it drowns in that enormous amount of dark data. ”
“ At our customers, we talk a lot with IT, and they have a lot of attention for data. However, the business does not need data, they need information. And too little attention is paid to that. Just look at the Internet of Things and all sensors that produce huge amounts of data. That data is often only briefly relevant.”
“If a sensor in a production line reports that something is broken, that is very relevant. But all messages of the 24 hours before, in which the sensor reports the part is still working properly, you don’t have to save. The data architectures are often barely equipped to carry out such balloting when storing data. And as a result, you see that it is still difficult for many companies to find a good ROI for new technologies such as Big Data and AI. The basics are often simply not in order.”
Business result
“We always advise clients to first thoroughly understand what the new technology actually does and to determine what business result they want to achieve. In recent years, IT has delivered many BI solutions, while the actual need of the market lies much more in predictive and preventive analytics. The techniques for extracting more predictive information from data exist, but the data storage is often not yet designed to be able to use it.”
“We also see that people tend to transfer things that are very complex to AI. let. But that is not by definition the sweet spot of AI at all. In the US, for example, an attempt is being made to teach a robot the skills of a brain surgeon. Billions of dollars have already been invested into that, while that robot has only 4% of the skills of a surgeon. Wouldn’t it be better to let go of teaching skills to that robot and focus on, for example, the use of AI to automate the surgeon’s memory?”
Robots
“And by that I mean: using AI to store everything that such a surgeon has seen and experienced in recent years regarding operations and disorders in an easily searchable database, on the basis of which you can perform diagnostic analyses, for example. A computer is much better at remembering things than a human being, while I think a human is much better at performing precision operations. In other words: let people do what people are good at and use robots and AI for things a computer is better at.”
This means that data and AI must be invested at the right level with companies. However, that is by no means always the case. “Our research shows that data management in at least half of the companies is entrusted to the IT manager. In the other half of the companies, new positions have often been created, such as an architect or an information manager. We also see that the quality of the data is less at those companies, where the IT manager is responsible for the data.”
Experiment
He concludes: “Companies that have shown the first actual successful implementations of AI and really innovate are the organizations that dare to experiment. Essent is a good example. I spoke to them about their participation in the benchmark. They carry out many pilots and pilot projects. 8 out of 10 of them fail, but 2 out of 10 are a success. Many companies wait to adopt a new technology until the business case and ROI are fully clear at the outset. But if everyone did that, there would never be innovation. And the promise of AI techniques is huge in my opinion, so it’s good that there are companies that dare to stick out their necks.”