Data is data and facts
Let's say we have a task: to draw up a work schedule for call center operators, which works 24 hours a day.
It is clear that the number of incoming calls at night is several times less than during the day, so fewer employees need to be replaced. However, this is not the only factor that needs to be taken into account. What else will need to be entered into the system:
- The average number of incoming calls on each of the days of the week, months, seasons (in summer, for example, they will be less due to the time of holidays).
- Distribution of calls throughout the day.
- The number of operators, their vacation schedule, established by law 2 days off per week.
- The geography of the location of the customers of this technical support service (for example, when it is night in Moscow, in Vladivostok it is day)
Other data such as employee salaries.
Self-learning systems allow not only to mechanically build a graph, but to take into account all data changes and regularly optimize it.
This is the ideal situation that the large digital banks that are building the ecosystem are striving for today. It will be some time before the all-in-one service becomes mainstream. Is this good for the user? Unless you're a conspiracy theorist, the answer is yes. The banking application is gradually turning from a rather narrowly functional service into a full-fledged universal assistant for all occasions, which takes into account the tastes and interests of the client. At the same time, the collected data science agency is impersonal and does not bear the risk of using it for incorrect purposes.
AI helps banks build more accurate and unbiased scoring models. Scoring is a system for assessing the reliability of a borrower based on a personal credit history. Algorithms not only check the client, but also make predictions about his future behavior based on data about other borrowers with similar characteristics. Earlier, at the dawn of the heyday of consumer lending in the mid-2000s, banks issued loans to everyone, but after the crises of 2008 and 2014 and non-payments, they began to build scoring models based on the accumulated history. As a result, the percentage of refusals has been steadily growing for several years, and the number of problem clients is decreasing accordingly. Today this process has begun to be automated. Sberbank already makes 100% of such decisions with the help of AI. Human intervention is required only in 5% of cases.
How it can help business
Directly, and for the benefit of both the seller and the buyer. The simplest and most striking example is targeted advertising in search engines. Google is based not only on queries in the search bar, but also on the user's movements (for example, seeing that a person visits Starbucks often, starts showing ads for coffee and coffee shops), his online communication, activity on social networks, etc.
Many companies successfully use data science consulting in their activities.
Netflix, based on the analysis of the behavior of its subscribers, recommends them an individual selection of potentially interesting films and series, and the "cover" of the content is created for each viewer personally.