The fad coming from the USA that will force the incorporation of AI in the process Surely it is only recently that we have started to hear a new concept in [...]Read More »
Artificial intelligence (AI) has gone from being a sci-fi movie theme to an extremely important topic in the business world. That's why today we bring you 2 case studies on artificial intelligence with a focus on sales and risk. so you can see how it works.
From predictive analytics to machine learning, the potential of artificial intelligence to support data-driven business decisions and strategies in different industries should not be underestimated.
Thanks to artificial intelligence is increasing our work, improving customer experiences, increasing productivity, efficiency and creating new sources of revenue. Here are 2 cases that Fernando Pavón, CEO of Gamco and expert in Artificial Intelligence applied to business, recently explained in one of the most recent AceleraPYMES cycle
When looking for a new AI-driven solution to a given problem in a company it is useful to have examples of companies that have done it before, an idea of how they developed the solution, who helped in the process and what the impact was or is. Below we leave you with 2 scenarios:
The first artificial intelligence case study is a sales topic. First let's look at what the application is about, what it achieves, and why it was necessary to implement this solution in a very small company.
In this case there are different user profiles of Purchase from the Southa marketplace that has been around for a short time and wants to increase the number of merchant affiliates it attracts.
The problem with attracting new customers is that we apparently still have no known data. However, there are millions of data at our fingertips thanks to internet services are giving. For example, Google Maps is giving us data from all the stores and segmentations. What we need to do is for the machine to incorporate that information.
To find new customers, there are multiple databases: I can buy databases from a third-party company that will give it to me; I can go to Google Maps and start looking and pulling up information and sweeping up customers from a certain zip code, etc.
Once I have this list of customers, their characteristics, telephone numbers, socioeconomic information and I know where they are positioned, the tool will be keeping a feedback: if they are interested, if they have already called, the actions they have, etc.ç
This application is designed to optimize the actions of telesales and also of salespeople on the street. When my salespeople sit in front of the system, they can see the list of customers, which ones they have to call, which is the source database for each one, etc. And, after each action, they will have feedback that will continue to guide them.
For example, when we have a large list of customers to call, our salespeople have two options: either they start calling randomly, or they do so in an orderly fashion based on the criteria analyzed by the tool.
In this sense, if we have a list of 1,500 contacts to call and we do it randomly, we will have to call them all and a conversion rate of 5%. If, on the other hand, we follow the indications of the tool, which detects those who are interested, we would only make 380 calls with a conversion rate of 20%. In other words, out of every 10 calls, I get 2 attentions. That's 15 points of difference with far fewer calls, so the costs are significantly reduced.
The second case study of artificial intelligence is based on clearly measuring the return and risk of non-payment. At GAMCO, we have a packaged solution: ARM, a very simple SaaS software where only with the invoices the system can see how many invoices are due and how much time they are overdue. The system marks a speed of mild, serious or critical non-payment.
The client that is paying me today, that I am giving them a service, I am billing them and logically what I want to do is to see if that client is deteriorating. Another thing that is observed is that the client does not stop paying overnight, unless things like confinement, covid, etc. happen. it is normal that there is a previous behavior that reflects that it will be a risky customer and may generate defaults..
ARM shows, within the list of customers, which are the most likely to generate a problem of non-payment and at a higher cost. In this way we will have information on which customers are those customers, which invoice has not been paid and how serious the non-payment is.
These alerts are issued early enough to take the necessary actions. Generally, we are between 75 and 80 percent detection of future non-payments.
There are three ways to integrate the data into the tool:
This will depend on the company, its maturity, the IT capacity it has, etc.
In addition, each client has its own instance with its data encrypted on hard disks so that only that client knows about that information. When people say that artificial intelligence has information confidentiality problems, they forget one very important thing: the machine does not need to see the data in plain text, but it can work with encrypted data and only when I am going to use it, it is decrypted to work with it.
Therefore, the information is secured and encrypted to be used to train the systems and for them to learn, but not that numerical data.
GAMCO's forecasts point to an increase of at least 10% in the percentage of "non-performing loans" to individuals over the next five years.Read More »