The credit recovery analysis process is now used not only by Finsolutia’s teams but also personalised for the company’s clients, allowing it to be used by several users with different profiles. This flexibility makes the platform even more valuable.  

In this process, credit recovery agents contact debtors in an attempt to obtain a payment agreement or a proposal to negotiate the defaulted credit agreement.  

This contact is usually established via a telephone call. In the recent past, users of the Finsolutia platform wasted much time when they hung up, as they had to summarise the call and enter it into the system.  

Although the average call lasts around six minutes, some can last up to 19 or 25 minutes. Due to the complexity of these calls, the process required platform users to remember the entire conversation and the information shared when drawing up the summary, often resulting in the use of acronyms and a lack of standardisation in the records. 

Finsolutia has “a system that contains all the information about credit management, so these notes are very important for future negotiations. One of the employees’ complaints was that they spent too much time summarising the call, i.e., they spent around 10 to 15 minutes summarising the call after hanging up,” explains Miguel Madeira, Executive Managing Director. He added that this time spent summarising the call represented a cost of around four euros/employee to the company.  

The idea was then born to use an Artificial Intelligence solution and, in a joint project with BI4ALL, to develop a process to automate this entire task and guarantee greater consistency and standardisation in the records.  

Although the negotiation is always made via telephone calls, a fully automated process was implemented, with the call being recorded. This process transcribes the call, analyses the sentiment of both the recoverer and the debtor and, in the end, through a Large Language Model, summarises the call and automatically records it in the company’s CRM. 

“For every call made by a loan recovery agent, they now save around 10 to 15 minutes, which translates into a huge increase in productivity for the teams, and it should be noted that Finsolutia has around 80 recoverers in Lisbon and Madrid doing this activity every day. In practice, our recoverers will make more calls. They won’t recover more debt, but they’ll recover it faster because they’ll have more time to negotiate,” explains Miguel Madeira.   

At the moment, the focus is on credit recovery, the company’s core business, where it wants to see a greater increase in productivity. Finsolutia also identifies other business areas, namely credit intermediation, where the company wants to start using these tools to record all the activity involved in attracting new loans for credit intermediaries.  

“We need to broaden the user base, which at the moment is around 80 users of this tool, and we want to extend it to the entire credit intermediation platform,” the responsible adds. 

Another important indicator is the level of satisfaction of both Finsolutia agents and debtors. “By analysing sentiment, it’s also possible to understand the performance and how each credit recovery agent works and carries out their duties. In fact, over the next year, we’re planning to implement a gamification system in which the agents with the best scores in the sentiment analysis part can get a bonus in their salary, for example. It’s a way of incentivising the use of the tool and increasing the quality of the teams’ performance.” 

In the future, depending on the business area, Miguel Madeira emphasises that it will be necessary to improve the prompts they use in Artificial Intelligence models. In addition, “we will also want to test the new models that are being launched, change the prompts that are more appropriate for the business areas, and then there will be some development, research, and testing to see if the results improve compared to what we currently have.”