What difficulties there will be when implementing artificial intelligence?
Analytical company Marketing Logic, together with a large bank was engaged in the introduction of artificial intelligence initially in marketing processes, then in network management and finally, in the selection of staff a few years ago. This case has become one of the most impressive in Russia, but in fact the team faced a lot of difficulties.
Implementation of artificial intelligence and problems that appear afterwards
Connecting AI causes different reactions among employees of the company, as it concerns all levels, from top management to ordinary employees. Most often people in bayonets perceive the introduction of new technology. The staff does not understand why the machine finds in its work positive or negative indicators, how it calculated patterns, by what factors. There is rejection due to the misunderstanding of how artificial intelligence works. Developing the technology, the company hired psychologists to understand how to accustom staff to machines. It turned out that the AI has to earn its reputation, and for this, colleagues need to understand how it works. Of course, it is impossible to turn all employees into developers and specialists in new technologies, so in parallel with the work on artificial intelligence itself, the company created an “assistant” who explained the decisions taken using language that is understandable to people. The worst scenario is the introduction of artificial intelligence, when the team does not understand it and is not ready to accept it. it is impossible to change at one moment traditions that have been developed for years and experienced employees will not suddenly begin to obey the machine. Therefore, the company thought out a smooth system of implementation.
The first year was marked as a “system of recommendations” that worked not instead of people but as a help for them.
It is logical that artificial intelligence should learn from someone and on something. This question had to be figured out too. At first, the developers believed that the system needs to “feed” the data for the last couple of years before the introduction of new technologies.
Then they decided that it is best to use the practice of the most successful employees as an example. But when the recommendation system has been working for several years, the following situation arises:
● technology has already gone through several learning iterations, and the tips have become “smarter”;
● staff can no longer show “best practices”;
● workers are increasingly accepting AI proposals;
● the team relaxed and emotionally “accepted” innovation.
Here comes the problem of shifting the selection. New information to continue machine learning does not arrive; the system becomes closed and can miss something. Usually this issue is solved by when a zone is allocated where employees should be against AI. But it is gradually becoming apparent that it is increasingly difficult for them to create and implement successful practices. Therefore, in such cases it is necessary to create separate groups of workers who would be experts in the question, but continued to look for new opportunities.
Another important point in the introduction of a smart system is that people see it as an opponent, and it isn’t a coincidence. Automation sooner or later leads to the fact that AI makes decisions better than even the most successful employees, thus saving company money. Naturally, the employees worry that they will be axed. This is an inevitable reality, and it is necessary to explain to the team clearly what they need to learn new skills and knowledge to work with machines, not against them.