Most read articles
|1||Why managers need to understand AI|
|2||Are Lawyers Risks and Crisis Managers?|
|3||Will a cashless society soon become reality in Switzerland?|
This blog follows the article “Do you already invest in AI projects?”. The previous blog clarifies important vocabulary on Machine Learning and Artificial Intelligence, which we will refer to in this article.
Firstly, it is important to mention, that data systems are only “cultivated” properly if the management is aware of the importance of this task, furthers appropriate means to it and cooperates proactively with specialist teams. There is nothing more frustrating for data specialists than working in an environment where the management does not support their activities or make unrealistic demands. Data specialists will then soon lose their motivation and react with cynicism, which can hardly be the key to success.
Executive School Programmes:
Secondly, the fact that algorithms recognize patterns, which humans often cannot grasp and interpret, results in a humongous ethical responsibility that programmers but ultimately also managers have to take. If there are algorithms in the world that make ethically questionable decisions because they have learned it from your (unmaintained) data, it is often very time-consuming to correct the damage. Only with a somewhat deeper understanding of how AI works, managers can be sufficiently sensitized to warning signals.
Thirdly, algorithms usually don’t do error-free work, even if they – hopefully – make fewer errors than the people they are working for. Let us return, for example, to creditworthiness. In the example of automated lending, two kinds of mistakes can be made:
- you grant a loan, but a default occurs; or
- you do not grant a loan even though the borrower would have repaid it.
The wonderful thing about algorithms is that they can predict whether you are more risk-minded and would rather make more mistakes of the first type, or whether you prefer to act more cautiously and therefore make more mistakes of the second type. If you are too cautious and hardly grant any loans, you can shut your bank immediately. It is evident that decisions concerning risks have more to do with a business strategy than with pure programming, as the choice is reflected in your business result! What you certainly do not want is to leave this decision to your team of programmers. Unfortunately, this is exactly what happens in many cases. Usually mistakes like this are happening within companies where AI is an unknown subject to the management. However, only if a manager is aware of this topic he will be able to work productively with his specialists and set the parameters together in such a way that they fit the business` strategy. In general, specialists will take you seriously as long as they realize that a basic understanding of AI is existing. This is very important to avoid cynical behavior of the programmers, which is unfortunately seen quite often in practice.
Logic behind the planning of data projects
It can be observed quite frequently that managers struggle with the “nature” of AI projects, as they are very volatile. If there are difficulties with an AI project, a higher resource input often does not lead to more success. This wrong assumption usually evolves from experiences with classic IT projects such as the development of new accounting software. Sometimes data simply does not cooperate, and this can already be clear after only three weeks. In such a case, it is often best to bury a project quickly.
From our experience, we can tell that managers tend to be struggling when it comes to handling this logic, especially when they do not understand why data is so volatile. Data needs trial and error, even more than an agile approach. If, for example, it turns out after a few days that the target variable – the “Y-variable” – is missing, the project is usually right at the end. At best, you can make the decision to capture this size from now on and start again in a few months or years. For some managers this is irritating, because the assumption that the existence of a large amount of data should be sufficient to make it possible to gain knowledge and create value is predominant.
What kind of project should you start with?
For your first experiences with AI, we recommend you to find a project that does not affect core processes, has low complexity, but a relatively high impact. Your first AI project should definitely be successful, because it has a lighthouse effect! An impressive example is customer behavior or the effectiveness of marketing campaigns. Picture recognition is also often quite easy to implement because technology components are already available.
A (real) example from practice is the following: A health insurance company sent 60,000 letters per year to convince customers to choose a higher insurance cover. On average, this marketing campaign was successful for 300 people, which corresponds to a rate of 0.5%. It was easy to use machine learning algorithms to filter those people, who were very willing to take out better insurance. Thus, the insurance company could only specifically write to about 300 people and the success rate was at least 90%! Approximately 59,700 letters no longer had to be sent and therefore did not end up in the trash.
To sum up, we are convinced that there are important reasons why managers themselves should have a basic understanding of AI. In particular, this allows a productive cooperation with the data specialists. Briefly, we see the following success factors and obstacles in the implementation of AI projects:
- Management has a basic understanding of AI.
- Data is “cultivated”.
- Data specialists make strategic decisions together with management.
- They start with a manageable project with great impact.
Obstacles- The idea that…
- more resources lead to more success.
- There is always something that can be done with sheer amounts of data anyway.
- you can leave everything to the experts.
- you don’t need to understand anything anymore when buying a system and can simply press the button.
- a large project is more successful than a smaller one.
Do you prefer watching a video than reading this blog? We then recommend our Webinar.
picture from iStock