Modern business administration: AI can be treacherous for entrepreneurs

Artificial intelligence (AI) is on everyone's lips.

There is a real gold rush atmosphere around data-based algorithms.

The possibilities seem almost unlimited, so it is not surprising that people are asking whether business decisions can also be automated.

Owners and executives of companies ask themselves three questions in particular: In which areas can business decisions be replaced by algorithms?

Which areas cannot be automated with the help of algorithms, and where are they actually harmful?

And how is the role of the entrepreneurial decision-maker changing in the age of AI?

An algorithm is a prediction mechanism.

It is based on correlations found in existing historical data.

For example, if I am a customer looking for a specific product, the algorithm tells the seller which other products I am interested in.

This prediction is based on which other products the seller's previous customers who were also looking for the product I was looking for were interested in.

A pattern is formed from this data, which is applied to my search.

The pattern can be refined as desired by including further data, for example about me or the situation of my search.

Regardless of the extent of the data involved, the basis of the prediction always remains a correlation.

What it lacks is causality, i.e. a principle of causation.

Tesla autopilot missed trucks

The quality of the prediction depends on the truthfulness of the old and new data and on whether the correlation also really exists for the new data.

The extracted pattern is not a proven knowledge, but an assumption that may or may not be true.

If it does not apply, it is a pseudo pattern.

For example, if a customer buys a book about growing roses, they will be served advertising on the subject even though they don't have a garden and the book was a gift for a friend.

The purchase was therefore only associated with the rose cultivation, but not causally dependent on it.

Such a mock pattern had deadly consequences for a Tesla driver whose driverless car crashed into a truck.

The Tesla algorithm actually interpreted the white sidewall of the truck as sky according to the mock pattern "white area means sky".

Algorithms can be used profitably for companies if it is not important not to know the causality.

This is the case when it comes to frequently recurring applications with stable connections.

An algorithm is much better able than a human decision maker to find the best place for an object in a central warehouse with thousands of individual products.

Other typical successful areas of application for algorithms are the purchase of consumer goods, the maintenance of machine parts and the planning of goods flows, but only under stable conditions, as the current chip crisis impressively shows.

If the conditions are not stable, the algorithmic control has fatal consequences.