SOLUTION

Credit Scoring Cognitive Automation

Virtual cognitive assistant for automation and artificial intelligence.
IBM Watson and IBM ODM on Cloud

Habber Tec has an artificial intelligence solution based on the services of IBM Watson and IBM ODM on Cloud (Operational Decision Manager) and IBM Cloud Native Development Subscription for IBM Cloud, specialized in the granting of real estate loans, this solution allows to improve the level of services provided to customers and thus support the business decision making. It is developed through predictive models of machine learning, both for the concession and detection of possible defaults, all this combined with business rules (risk policies) and the use of cognitive services which improve the user experience and the capture of new data. This solution is 100% based on SaaS services, equipped with a dashboard that allows the tracking and monitoring of KPIs defined to the platform. It is a fully scalable solution.

CLIENT

UCI – UNIÓN DE CRÉDITOS INMOBILIARIOS

UCI had the challenge of quickly answering the clients that requires mortgages for the more than 700 real estates for which it works, as well as its direct channel hipotecas.com. They opted for the Habber Tec solution that not only allows them to automate their risk decisions, but also learn from the experience of previous operations.

The Habber Tec solution uses artificial intelligence from IBM Watson, it has been developed with Watson Studio and deployed in Watson Machine Learning for its execution, also combines AI algorithms with risk policies using IBM ODM on Cloud, which has allowed the creation of a business-rules-system adapted to the UCI risk department, providing such department with a versatile and scalable solution where they themselves are able to set the parameters to apply for the granting of mortgages, these rules being applied in real time.

Using the history of UCI mortgage concessions, the Model Decision has been trained, responsible for deciding whether a mortgage should be granted or not (as an experienced bank agent would do) and the reasons for the yes or no so the artificial intelligence does not act as a black box, but is interpretable and capable of complying with the regulations of the GDPR.

This has resulted in a powerful model that allows to provide with the minimum human intervention an instant response of 82% of credit applications, with an accuracy greater than 84%.

Additionally, and using the UCI payment history, the default model has been trained, which predicts with an accuracy greater than 80% the probability that the mortgage will be unpaid, based on the profile of the applicants.

Thanks to the machine learning capabilities offered by Watson Machine Learning, the system can retrain with new decisions being able to adapt automatically to the constant changes that the sector is experiencing.

BUSINESS PROBLEMS

From the beginning, UCI has carried out the credit analysis process through the real estate professionals of the company itself. With the constant technological evolution, UCI has decided to enter the digital transformation, since it did the process manually and only depending on the professionals, it generated the following consequences:

Delayed response: they invested time in the process of reviewing a mortgage loan application to approve or reject it (it took days).

Loss of competitiveness: because they did not provide agile service (real-time responses) to their potential clients, they missed the market opportunity.

Presence of subjectivity: the final decision is made only by a team of specialized people, with feelings, prejudices; that, even if they stick to the rules, they are subjective and that may involve decisions that were not so neutral / systematic / objective, in some cases.

Difficulty to adapt systemically to the constant changes in market-based risk policies.

Non-payment on granted mortgages: Many of the approved mortgages were not paid.

BUSINESS OPPORTUNITY

Given the situation that UCI has carried out the credit analysis process through professionals of the company itself, the creation of a solution capable of evaluating real-time applications to prequalify potential clients has found a balance between the number of approved mortgages and the number of defaults. This solution gives UCI a great opportunity to continue growing in its business and, above all, improve its results.

TECHNICAL DETAILS

The solution is structured in 6 stages:

1 Leads approved by the basic filters.

2 Data reported by the client (form, without guarantees, such as demographics, income, work history, AIS, property data).

3 Machine Learning.

4 Mandatory rules (max. 40% over the incomes).

5 Pre-decision (Yes, No, Check).

6 Decision details:
What motivated the decision (data with positive, negative or neutral weights).
Pre-qualify based on the policy (if the decision is NO, what must change to get it)

IMPLANTED IBM TECHNOLOGIES

IBM Data&AI

  • Watson Studio
  • Watson Machine Learning
  • DB2 Flex

IBM Cloud

  • COS (Cloud Object Storage)
  • CloudFoundry (SDK for NodeJS)

IBM Automation SaaS

  • ODM on cloud

IBM Cloud Native Development Subscription for IBM Cloud

COMPETITIVE ADVANTAGE

Reduction of the time and resources necessary to manage business files and processes.

Improvement in transaction costs per file.

Time2Market.

Productivity increase.

Reduction of mistakes and fraud.

Improves communication with the client.

Outsourcing of non-core activities.

Possibility of analytics.

RETURN ON INVESTMENT

Save on personnel costs. What several people used to do, now a machine does.

Increase in income, because mortgages are being approved faster.

Reduction in defaults on mortgages, since those requested are more efficient and accurate.

CONCLUSIONS

It is important to note that any company which works with credit approval in any market can use this solution to automatically pre-approve the credit and with a minimum margin of error that represents a decrease in credit default.

VIDEO

The following video explains how Habber Tec adapted its artificial intelligence solution with IBM Watson for Unión de Créditos Inmobiliarios that improves the level of services provided to customers and supports business decision-making. Through predictive models of machine learning, combined with risk policies, cognitive services that improve the user experience and capture new data, this application can give a real-time response to mortgage applications, automating 80% of decisions of credit.