For financial services technology companies like Fiserv, AI is the golden child.
“The company has been bullish over-investing in AI and analytics. We are building analytics capabilities in house rather than focusing only on services or an advisory play. We have developed other important technology models with in-house capabilities. We use AI to monitor transaction data in real-time and detect and prevent the occurrence of payment frauds. Similarly, a ML-based analytical solution has made it possible to create better personalization with recommendations and offers based on customer segments and transaction behavior,” Srini Krish, President, Fiserv Global Services told ETCIO.
Leveraging AI & Analytics to strengthen internal teams and operations
Fiserv is applying analytics for decision making. This includes amplified usage of analytics to gather actionable sales and marketing intelligence on visitor segmentation on mobile and web-based banking applications.
“Some evolving areas include the use of Natural Language Programming (NLP) for text analysis of customer service requests and surveys and intelligent automation like leveraging Intelligent and Optical Character Recognition (ICR/OCR) for cheque fraud or cards disputes. We have ML-based analytics which helps in rescuing false alarms and identifying true ATM service failures, thus decreasing service costs and improving customer experience. Some other areas where we are applying AI and ML include using fuzzy logic to find related counterparties in transactions to measure risk exposure and detect fraudulent behavior,” Krish added.
Krish’s team has built data models with in-house capabilities to analyse transaction channels, and address challenges faced with several uses. To develop and deploy analytical solutions, Fiserv leverage in-house platforms, open-source tools, and proprietary software.
“We are creating assets in the form of frameworks, codes, model validation programs as well as for analytics solutions. The intent is to break down an opportunity into modular forms – such as pre-processing, modeling and post-processing for AI/ML applications. These modules are standardized and automated to reduce time and run iterations to find the best in class models from an array of algorithms. We strive to make our AI/ML models reproducible, portable and deployable – productionizing the model is crucial for any AI/ML use case, along with continuous improvement of the modeling framework,” Krish emphasised.
Data and analytics-based decision making has emerged as the norm and represents a significant opportunity to optimize operations and break down decision-making silos. This proactive data-based decision making is made possible by timely actionable insights, which can now be achieved using AI and ML.
While the technology brings the spirit of innovation, the challenge of managing biases associated with it becomes even more critical.
Underlying data the main source of biases
“Biases can also be introduced in the process of collecting or selecting data for use. This can be at least partially addressed with “human-in-the-loop” decision making, which involves algorithms providing recommendations or options, which humans double-check or choose from. In such systems, transparency about the algorithm’s confidence in its recommendation can help understand how much weight to give it,” Krish highlighted.
According to him, to address AI-based biases, we need to focus on AI solutions that enable explainable and predictable decision making. These solutions should be integrated with human processes for appropriate oversight.
“While deploying AI, we use a portfolio of technical tools and establish responsible processes and operational practices such as “red teams”, a team that consists of security professionals who identify vulnerabilities in the system and test security postures; so that AI can be efficiently leveraged and biases can be mitigated,” Krish added.
AI is in no hurry to leave its throne in 2021
“The emergence of AI platforms and AI on the cloud are going to create more opportunities. Techniques such as graph networks for fraud detection, intelligent automation for operations efficiency, digital analytics, and creative AI will enable organizations to enhance insights. This in turn elevates the quality of data-based decision-making and offers an opportunity to increase flexibility and scalability. The use of newer techniques such as the creation of a real-time CDP (customer data platform) or customer analytics hub is also expected to grow to help in measuring engagement, customer experience, and profitability. We will continue to invest and build these capabilities driving innovation as these technologies individually and when integrated will be foundational towards application and product modernization, digital transformation and business continuity,” Krish concluded.