A new article published by me on the platform Industry of Things.
Industrial companies are increasingly launching promising AI projects. But after the proof-of-concept, the process falters: important data is missing, the results are disappointing, or the concrete use case is not so clear after all. What you can do about it.
Living in today’s world
5:43 a.m.: The alarm clock rings because it registers a light sleep phase in the specified time period. We immediately reach for the cell phone, which unlocks thanks to facial recognition. A thought is dictated and saved directly as text with speech recognition. The smartphone reports a comment in Cyrillic to Instagram – one click is enough to read the greeting message translated into German. At the desk, the laptop displays the most important tasks: A good friend’s birthday is next week. Quickly go to Amazon and directly order one of the suggested items.
These examples show: Artificial Intelligence has arrived in the private sphere. But what about AI in companies? The experience of the last few months shows: Many companies find it surprisingly difficult to implement AI projects in practice. They often start with enthusiasm, but come to a standstill after the first steps. This is not due to technical hurdles, but primarily to a vague idea, insufficient or unreliable data, and a lack of vision for integrating the results into business processes.
Differences: private and business AI applications
But what’s the difference between consumer and business AI solutions? For home consumers, the use cases have the following in common:
- It’s about a small, unique problem.
- The application has an enormously large user base.
- Accordingly, there is a lot of diverse data and data related to this problem.
- The provider solves exactly one concrete problem with the AI in each case.
In industry, there are also numerous use cases such as predictive maintenance, chat bots or automation. But these are individual – and often encompass several problems. The quality of the data is extremely variable. They are often unstructured or isolated, and frequently not all the necessary information is available.
In addition, suitable contact persons and experts are missing – or they have other priorities. A lack of domain knowledge for interpreting and analyzing data also hinders projects. Integration into processes is difficult. There are also organizational hurdles such as departments working separately or unclear responsibilities. In addition, there is a lack of trust in decisions as well as false positive results of AI solutions in case of insufficient data basis, faulty training or unsuitable algorithms. However, companies can lay the foundation for their AI projects with the following success factors.
…
Read more within the full article (in German): KI-Projekte richtig starten
