AI and machine learning are expected to add trillions of dollars in business value in the years to come through increased revenue, improved product development, and operational efficiency. Investment in AI startups and technology is increasing rapidly. And yet, questions about how to use artificial intelligence, let alone what aspects should be implemented first, can leave financial services firms paralyzed while at the same time feeling the anxiety that they not be left behind.
What we can’t see, AI and ML sees in plain sight
The topic of defining AI alone could fill volumes, but it really depends on domain and context. For financial institutions, a working definition could flow from what AI does uniquely well and what can best positively impact a bank’s competitiveness in the marketplace, which is the power to make meaningful predictions. Predictions are inputs to decision-making.
This kind of AI is enabled by machine learning, which is a unique methodology and technical infrastructure that allows algorithms to work through vast amounts of data and find connections that the human brain would never be able to process. ML can “think” in many dimensions in parallel, processing millions of variables at once, where the human brain can only handle a few at a time at most.
This kind of processing allows AI to find new market segment opportunities hiding in your data sets, and allow you to sell the right product, to the right customer, at the right time, at the right frequency, and in the right manner. It also allows us to flip the servicing model on its head. Imagine if it was possible to predict that an ATM would require service in the next two days.
What you do now is what matters
So, how do we hop on this AI train? What are best practices that help me succeed quickly and more often than my competitors?
- Bust down your data silos. The key to creating value with AI, is the availability of unified data to deploy models and take action on the results of AI models. You must invest authority in your teams to define a shared data strategy, launch new technology infrastructure, establish data governance, and foster data citizenship. Find the people that currently work around your data silos and ask them what they need to be even more effective. Two hours spent with a data analyst in the trenches, truly listening to what data you have and don’t have, is potentially worth weeks with a management consultant.
- Be bold. Consider Jeff Bezos’ edict at Amazon that all new code had to be deployed as a service. That commitment led to a whole new culture and became what we now know as Amazon Web Services (AWS). Make a commitment to embed AI in every new financial product. Encourage curiosity about AI can create value in each user experience and how current technology limitations are impacting your vision for customer engagement. Require product managers to explain why AI and machine learning don’t apply to a new product or service.
- Create a proof-of-concept (POC) culture that moves your business forward, tests your hypothesis, and enhances your practice of embedding AI. POCs that test technology operate on too narrow a field, which is why they often fail. Instead, engage promising third parties to augment your current team and practice learning how to embed AI. This will act as a force multiplier for every dollar spent in innovation.
- Focus on making steady progress on big picture and data-driven business problems:
- How can you identify customers at greatest risk to default and take a clear action to intervene and understand their needs?
- How can you predict which customers, and potential prospects, are likely to be the most profitable and create personalized marketing engagement strategies to retain and grow your share of wallet?
- How do you find new segments of customers that need your products, but are otherwise difficult to find because the signals and relationships are complex and hidden in your data sets?
- How do you focus your follow up strategies for customers who are likely to attrite from a given product or your institution entirely?