Gartner expects artificial intelligence (AI) to create 2 million new jobs by 2025. AI and machine learning are already an important part of business processes and business areas in many companies and organisations, making everyday work easier, optimising interactions with customers, reliably predicting the failure of machines or supporting the care of networked patients.

So far so good. Expectations of what AI can deliver are virtually limitless, but they are not always in line with reality. In meetings over the past few months with customers and prospects, my colleagues and I have seen exciting, surprising and groundbreaking ideas and applications, but also often exaggerated expectations. Is it time for a reality check? I think so, and here are my nine reasons.

1. AI is not the answer but a building block of your future business model.

False promises of salvation, such as the misunderstanding surrounding so-called self-learning systems, have raised exaggerated expectations, which have often been disappointed. The technology is not science fiction. It is simply mathematics, used to control and improve operational processes. Companies still must create their own business models.

2. Machine learning cannot deliver miracles. A culture of fact-based decision making requires companies to work hard.

Methods such as machine learning are simply the ongoing development of established analytical methods. There are no mystic black-box effects. Anyone who has not already managed to integrate analytics into their development and decision processes will not be any more successful with machine learning.

3. The successful establishment of data-driven business processes does not depend on the number of data scientists, no matter how well qualified.

To work creatively with data, a much wider range of employees must understand the basics of statistics (such as correlation does not equal causality). Companies also need to ensure that they do not put their mathematicians in an ivory tower. Only the collaboration of experts and business users will deliver successful data-driven decisions and business models. Otherwise, companies will not be able to gain any real competitive advantage with AI.

4. AI cannot compensate for a modernisation backlog.

Implementing AI does not improve business processes that were not already optimal. It may even expose more problems. Processes need to be constantly evolving so that AI leads to innovation.

5. The best corporate culture for AI is one in which ideas fail quickly and “successfully.”

Trial and error ensures that you can quickly see which developments are going in the wrong direction and change them. Even several “failures” in a row should not cause employees and management to lose their motivation. This stamina is ultimately part of the secret recipe for true success. Companies need structures that allow new ideas and fast decisions for or against their implementation.

6. The biggest technological challenge is lack of governance because the newest algorithm is not necessarily the best.

Quality assurance  for analytical models is indispensable. Many companies experience falling performance because of systemic bottlenecks, poor interfaces, or lack of clear responsibilities.

7. Artificial intelligence needs a new platform economy in which governance and openness are not contradictory.

Contemporary analytical environments are no longer homogeneous. They can therefore now only be controlled via open platforms, controllable APIs and holistic metadata control.

8. Data protection is renegotiated.

The General Data Protection Regulation (GDPR) is currently seen as driving consistent data management. In the future, the trading of "data for services" will increasingly be negotiated in individual business relationships.

9. Ethics and regulation are geared to new circumstances.

In the future, we will all need a better understanding of how algorithms work in principle. It is vital to understand how to integrate ethical, moral and other principles. The transparency of a fixed set of rules will become less and less important, and interpretability more so. This requires a powerful analytical platform that can handle a lot of data and will allow users to try things and compare options.

AI and machine learning are rapidly becoming a business reality. They can be used in enterprises to optimise business processes, increase customer satisfaction and service quality, or improve productivity, but only with a powerful analytical platform. This is important to enable businesses to handle and manage a lot of data. But while AI is becoming a reality, it should not be the panacea for all corporate problems.

Andreas Becks’ blogpost was inspired by “How AI Will Change the Way We Make Decisions” by Ajay Agrawal, Joshua Gans and Avi Goldfarb. Find the original article, as well as more about other important aspects of AI, in the Harvard Business Review report “Risks and Rewards of Artificial Intelligence.”

Download your copy here.