Moving from AI awareness to meaningful implementation
While most executives at financial institutions agree that artificial intelligence (AI) is important to their organization’s success, few have fully implemented AI projects. In a recent Cognizant survey of 230 financial services executives, three-quarters said AI is extremely or very important to the success of their organizations. However, only 61% of those were aware of an AI project at their company. Even more telling, only 29% were aware of a project that had been fully implemented. Clearly, AI is quickly becoming a competitive requirement, creating the risk that those who are not implementing or updating AI capabilities will fall behind.
Though leaders at financial institutions realize they need to infuse AI into their organizations, it seems that, for the most part, they often do not fully understand the tangible value AI delivers. AI can enhance customer experience, improve trading and portfolio management, assess the creditworthiness of loan applicants and detect potential instances of fraud or misconduct.
AI is a journey
So how do executives move from thinking idealistically about AI to applying it in practice? The AI journey starts with understanding the current state of the organization and envisioning an AI-based future. The first and most critical factor is data. What data do you have, is it accurate and do you have enough? Is your data consumable, and can it be labeled?
For example, lenders have historically made loan-approval decisions using a statistics-based risk model, with humans making the ultimate decision to approve or deny a loan. AI can augment this highly manual process by optimizing the suggested decision based on desired outcomes. Humans ultimately make the call, especially in cases where not enough data exists for a pure AI decision.
In most cases, implementing AI does not require replacing existing systems. Instead, applying approaches like deep learning to unstructured data can augment existing statistical models and extract insights that help decision-making.
The best action to take is not always obvious, as most actions are measured against multiple outcomes. Because deep learning is batch-oriented, it can prove expensive and not easily repeatable. However, by adding approaches like reinforcement learning or surrogate-based decision making, we can improve adaptability by incorporating an update cycle as part of the process.
Still, incumbent models lack certainty in their predictions, making them risky to use in real-world situations. A deep learning model, for example, might predict with 90 percent certainty that a particular decision would make a lot of money. But deep learning is notoriously bad at knowing what it does not know. Consequently, the 10 percent error might result in a greater loss than the 90 percent upside. AI certainty models can be used effectively to calculate a rather accurate confidence measure for predictions from other AI models.
One of the more powerful AI technologies currently emerging is evolutionary AI (EAI). EAI can automatically design deep learning systems without human programing, which helps alleviate the challenge of accessing talent and lessens the time needed for deployment compared with other technologies.
Using EAI, you can create a model based on observing historical decision data to come up with a decision strategy. It then uses that strategy, observes the results and repeats the loop, iteratively incorporating new observations to modify and improve the decision strategy.
Financial institutions can also take a hybrid approach, which allows AI to automatically make decisions if it has certainty above a defined threshold. This is a win-win, as the likelihood of achieving the desired outcomes is high. However, if the AI falls below a certain threshold, humans need to review it and either accept, reject or modify the decision. Eventually as the AI experiences more, it will become more certain, allowing you to gradually reduce human involvement. In cases where the decision loop is too fast for humans, such as in high-frequency trading, AI must make all the decisions, based on objectives and outcomes set by humans.
Designing an AI strategy
To capitalize on AI’s ability to bring fundamental changes across your organization, an institution’s leadership needs to be on board. Crafting an effective strategy starts not by looking at the technology capabilities, but by deeply understanding the business needs and opportunities AI can address. The following seven guidelines will help get the process started:
• Cast a wide net. Conduct a comprehensive assessment of your business processes to identify opportunities to apply AI, and estimate the potential benefits and investment required.
• Go beyond insights. First identify the most impactful decision loops, and then assess where AI can augment manual decision-making.
• Look for opportunities to leverage data. Assess your data for accuracy and explore nontraditional data sources. Machine learning can help identify which types of data are most important for specific business outcomes.
• Enhance data management and governance. Financial institutions tend to struggle with legacy systems and siloed data, requiring an upgraded data architecture that can quickly deliver diverse types of data to AI applications.
• Prepare business processes for digitization. Optimize processes before applying AI through system changes, standardization and consolidation.
• Acquire AI expertise. Securing talent can be extremely challenging, which is why many financial institutions choose to partner with fintechs that bring AI expertise.
• Encourage experimentation and discipline. There are no turnkey or one-size-fits-all AI solutions, so managed experimentation is key.
Done right, AI will drive business value across your organization — enhancing customer experience, improving investment strategies and streamlining back-office operations. Approaching AI as a critical business strategy will position your organization to succeed in the real world of artificial intelligence.
Babak Hodjat, is the vice president of evolutionary AI at Cognizant, a professional services firm based in Teaneck, New Jersey, that focuses on transforming clients’ business, operating and technology models for the digital era.
Want more BAI Banking Strategies? Sign up for our free newsletter!