To be able to do that, it's not enough to take up AI and launch a pilot - often referred to as an accelerator - with a few experts, isolated from reality.
We need to come up with an idea of where artificial intelligence can provide useful assistance in day-to-day operations. To do this, you have to understand what makes this new "AI colleague" tick and what he needs to do his new job successfully.
An example of why this is so urgently needed? Lead management in sales: Before the sales employee wastes his time, the AI colleague should assess which leads are promising and worthwhile to work on, and give the signal at exactly the right moment - i.e. shortly before closing - that the sales person can collect the signature from the customer.
Managers are enthusiastic about this idea, experienced salespeople less so. For one thing, they know it's not that simple. On the other hand, they don't want to be patronized by technology.
It is not only the hard facts such as name, address, industry and telephone number that count. Human colleagues take into account, sometimes consciously, sometimes unconsciously, many other factors, relationship networks, previous contacts, current satisfaction with the service, experience with the products, the competitive situation, and so on.
The AI also makes use of corresponding data, provided it is available. The more granular, the better. It searches for patterns, calculates the behavior score and match score, and shows whether the investment in the contact is worthwhile or not.
For this, it additionally needs a framework in which it acts. Thus, the AI is not too different from a human colleague, but its perceptions are limited to the pure data level.
So the real challenge is not so much the AI itself, it is the data without which it cannot learn. It must be collected consistently and in a structured manner and then used in sales and service.
But this requires enough of them - without Big Data, there is no AI, because without differentiated patterns, no reliable conclusions can be drawn. But this also means that without CRM as a basis, nothing works in our example. Surprisingly, all this is not so new!
Today, however, the CRM system must be networked in order to aggregate customer-related data from personal contacts, ERP, web store, customer portal, website and various other contact points, the so-called touchpoints. Automatically?
Preferably yes. Because as soon as one employee is responsible for recording data completely, it becomes time-consuming and gaps can be foreseen.
So to hire an AI, you first need to understand what it can be used for and how to train it. But then the problems start:
AI "thinking patterns" are usually so complex and involve so much data and patterns that it is hard to understand how decisions are arrived at.
So if the sales department is also told why the AI decided the way it did: Jackpot! Most of the time, however, this remains a mystery to the human colleague.
Artificial intelligence is therefore not a miracle cure either, but is based on things we have known and know for a long time. Its recommendations are more human and error-prone than often assumed or hoped for.
As of today, AIs offer assisted rather than autonomous driving. They can be our CoBots that support us and that we switch on when needed. They help in everyday life, take over tedious jobs and then hand them over to the true professional to make decisions.
However, he has previously defined exactly what he wants from the AI and can assess its impulses. But we should not underestimate these CoBots: In the future, they will also continue to gain autonomy in companies.
This is because AIs come up against limits above all as long as they do not trade directly with each other. Wherever their algorithms can connect directly with each other, they can make valid decisions under clear framework conditions.