Mike leads consumer lending business for a global financial organization. He is looking to upgrade its existing systems from a legacy platform to an AI-based solution. He has heard about the power of AI and how it can help businesses. Within his organization, Mike can see its applicability in risk models and also in taking some pro-active action to check whether the system can predict customer delinquency. As Mike explores this option further, he also understands why existing risk segmentation models are rudimentary based on static rules or statistical models. However, should Mike need to change his core collection and servicing platform to get benefits of technology? He fears that a rip and replace of collection systems would take too long and his business would get impacted if he is not able to use the power of AI sooner. After double-clicking the initiative to implement AI in-house on top of the existing collection system, Mike starts to lose hope in getting faster results. He finds out, that manual data preparation, feature engineering, and model building is a time-consuming process. Also, productionizing and realizing business outcome is a massive challenge with in-house AI implementation. However, Mike also has other problems to tackle. In spite of technology coming on leaps and bounds over the past decade, AI remains a black box, and this presents a significant challenge for audit and regulatory compliance. Although Mike can choose to be guided by an AI system, if he cannot defend or explain the recommendations it makes, and this would place him in a vulnerable position in front of auditors and model risk validators. How can he balance what the business needs faster with what regulation and compliance demand? If you identify with Mike’s conundrum, read on.
Fast becoming the cornerstone of digital transformation, the proliferation of AI has evolved from automating repeatable tasks to powering human-led decision making. Its significant advantages, however, are weighed down by some very real challenges – compliance, traceability, and auditability. This absence of transparency does little to inspire trust in AI-powered decisions, especially when the human being taking the final call accepts liability for any errors. In the financial industry, the need for organizations to make regular judgments about customers exacerbates the problem. Companies need to be able to explain and back the decisions they make, while customers need to understand the systems assessing them, neither of which is possible with conventional cognitive applications. In the financial sector, from fraud detection to lending, we have seen that the hurdle of transparency offsets the appetite for AI adoption.
The alternative, however, is hardly ideal. Consider the lending sector; Data models are updated no more than twice each year, generating dated results may not correspond with the actions eventually taken on consumer accounts. They also take into account a limited number of criteria before creating a broad and inaccurate borrower risk profile. This traditional output from legacy systems offers sub-optimal segmentation, preventing businesses from understanding the risk profile of their delinquent portfolio and, more importantly, the true recoverability of overdue accounts. Intelligent systems could transform these operations by introducing efficiency and accuracy that saves millions of dollars while increasing customer satisfaction. The answer lies in a solution that offers the transformative power of AI while providing human-level contextual explainability.
That’s why we built CollectEdge. A thin layer of AI-powered intelligence that sits on existing core platforms, CollectEdge is a powerful solution that integrates smartness, transparency, and accountability into existing collection systems. CollectEdge helps organizations reduce delinquency rates, improve recoveries, and be more operationally efficient through actionable business insights backed by clear explanations. The solution offers a substantial advantage over existing systems, and its core advantage moves well beyond efficiency. Here’s why.
When you make a machine learning model, the objective is to ignore correlated information and create a feature set that is a combination of two to three data points. We understood that it was essential to develop a system that could generate smarter data models, interpret their output, and offer an easily accessible narrative. That’s why with CollectEdge we built a model that tells you the decision and then shows you which influencing factors helped arrive at the result. These insights are not available in a conventional analytical or augmented analytical system, but are offered by CollectEdge because of its ability to provide decision-level explainability. More importantly, it gives businesses a thorough and explainable understanding of the influencing factors, the reason behind their selection, and their impact on the final recommendation. Companies can also identify and utilize complex links between multiple influencing factors to inform upstream processes, future models, and collection strategies.
Every prediction that CollectEdge makes is derived using specific influencing factors. It’s important to note that these are not just analytics-driven insights. By using a wide variety of influencing factors including, for instance, macroeconomic indicators, CollectEdge offers end users powerful predictive intelligence based on a comprehensive view of the data subject. Combined with the solution’s explainable AI, this ability means that users can now see the thinking and data points behind each recommendation allowing for better auditability, compliance, and regulatory discipline. Additionally, if there is data available outside of the current base product model, it can easily be added using CollectEdge without having to repeat the model creation process.
We believe this eliminates the need for organizations to choose between a less elaborate method with a high level of explainability and a sophisticated tool with little or no transparency. To make it easier for financial institutions to adopt AI, we integrated explainability into CollectEdge in a way that feels intuitive and actionable to end users. While AI-based systems are usually a tradeoff between accuracy and clarity, CollectEdge’s model-agnostic approach provides enhanced explanations. These are provided at the local level instead of model-level recommendations, allowing teams to extract actionable business insights that can be fed upstream into the process and, in turn, deliver real downstream impact.
The best part of the solution is that it can deliver sizable improvement in business performance, including reduced charge-offs, savings on call investments, improved CSAT scores, and near real-time models. To allay any implementation concerns about the magnitude of effort or investment, CollectEdge sits unobtrusively on top of existing systems without any need for ‘rip and replace’ allowing organizations to experience the power of AI adoption in production.
With CollectEdge, companies can now adopt a future-proof AI-led decision-making process that their leadership, regulators, and customers can trust. This addition of efficiency, intelligence, and clarity frees up resources and investments that could drive significant business growth. Now that needs no explanation.
To experience the power of CollectEdge, schedule a demo today.