data silos

Just in time for Halloween, the horror scenario was finally coming to an end.

October 31 of all days was chosen to finally open Berlin Brandenburg Airport with a delay of nine years and additional costs of five billion.
It is not known whether the political elite appeared at the opening in disguise. On the other hand, the numerous faux pas that took place during construction are well documented.

And there is always something to learn from at least one of these numerous mistakes: Data silos will become very expensive at some point.

The planners, property developers and contractors on the construction site in Berlin-Schönefeld all worked with a wide variety of planning tools and software products. And each of them had its own data format. And so it happened, for example, that these “data inconsistencies,” as they were called in later reports, meant that the electrician had different plans than the fire protection planner.
As a result, pipes, and cables were laid where they did not belong at all.
There was no technical manager who had an eye on data homogenization. And even if he had existed, he would hardly have had anything to homogenize, as the various software manufacturers had not yet agreed on a data standard.
The GAU in construction is certainly a disaster that is second to none. However, it is a problem for many.

According to a study by management consultancy Accenture, 80 percent of all companies worldwide are struggling with data silos.

Large corporations use between 200 and 400 different software products. And SMEs are also equipped with a good two-digit number of software tools, depending on the industry. Almost no company is able to centrally combine the data and use it sensibly.
On the contrary: New software products are added every year, increasing complexity for 69 percent of companies and causing a high level of data inefficiency.
The shocking result: 70 percent of all data remains unused.

In a digitalized world where data is the new currency, this is a dilemma. In three out of four companies, data silos hinder internal collaboration. 74 percent of managers regard this as a competitive disadvantage. But that is just the tip of the iceberg. The real danger lurks beneath the surface.

Management consultancy Gartner predicts that 80 percent of all AI implementations at companies will fail if they do not develop a data strategy in advance and clean up their data.
According to a study by PwC, this means that 57 percent of companies are stuck with their AI efforts and unable to scale.

In order not to completely lose touch, this problem is leading companies to a half-hearted use of artificial intelligence. According to the motto: Let's try this out in one area and then see what effect the use of AI has on us.
As a rule, this leads to the realisation that the use of artificial intelligence only offers moderate advantages.

This approach conceals not only the data dilemma, but often also a lack of understanding of the technology.

‍
The emphasis in AI is on “intelligence.” You can't introduce this into a company like new software, but you have to build it up. Intelligence depends on connection and networking, the interfaces to all areas of the company.
Like neural connections in the brain, it only becomes intelligent when it has as many nodes as possible.
An isolated use of AI can therefore never develop its full potential. Artificial intelligence can be used gradually, but must always take a holistic approach.

The good news is that data silos can be broken up with technological solutions.
‍

In most cases, they are not a “technical problem,” but an expression of a lack of data culture.
To eliminate data silos, you first need a clear data strategy with central governance that defines responsibilities, standards, and data quality.
The data must then be technically consolidated and harmonized via central platforms such as data lakehouses or data fabrics.
Finally, consistent master data management ensures as well as automated data integration to ensure that data is available cleansed, up-to-date and AI-enabled.

‍

Editors: Thomas Satori | Mathias Keswani