Financial Services, and especially retail banking, has long been the beating heart of decision automation. Not only did many of the principles of effective decisioning get developed in financial services, many of the most powerful use cases are also found there.
Whether it’s automating fraud detection or loan origination, targeting offers or driving next best action marketing, performing Know Your Customer or anti-money laundering checks, the value of decision automation has proven itself time and again.
Let’s look at some of these use cases and see what they have in common.
Regulatory Compliance
In every case we have a high-volume transaction or real-time interaction where the right decision for handling it is not obvious. Financial institutions have millions of customers and accounts, and their customers transact on those accounts every day. Detecting and preventing fraud, preventing money laundering, approving payments and transfers between accounts require making good decisions at very high volume as a result. Loan and overdraft origination and other credit-risk decisions may be lower volume, but customers expect a real-time response, so automation is critical to retaining and developing customers. Complexity in these decisions – the reason the “right” decision is not trivial – comes from regulation, broad and deep datasets and the need to assess risk accurately.
Automating these decisions uses a classic combination of technologies and techniques.
Transparency
We need to apply decision modeling to make sure we understand the decision and can engage business SMEs in specifying how our decision must work. Decision modeling allows us to handle the complexity of these decisions by breaking decisions down into simpler, more manageable pieces.
Many of these sub-decisions are regulated or driven by published policies, product descriptions, contracts or standard practices. Using decision tables and other business rules to document this logic so that it can be explained to customers, shown to regulators and understood by business, legal and operations staff delivers essential transparency and agility.
We need to analyze broad and deep datasets to identify outliers and patterns of fraudulent behavior, to find predictors of credit risk, and to understand our customers’ journeys. Simple statistical analysis has evolved through data mining and predictive analytic tools to modern machine learning and AI platforms that deliver increased accuracy and respond more quickly to changes in our data. These predictions and classifications must be integrated into our decision-making and guided by the policy and regulatory rules we have outlined.
Dynamic Adaptability
Finally, the pace of change in financial services as well as the competitive market in which most institutions operate, means that continuous improvement is the order of the day. We can’t expect to leave a decision-making approach unchanged for any length of time. Which is why we generate detailed logs of our decision making, continually improving the rules and ML/AI models we are using and invest in champion/challenger and A/B testing,
The Future of Automated Decisioning within Financial Services
Recently, the very longevity of decision automation in financial services has led some to challenge the necessity of decision automation – do we REALLY need to keep using these established systems, technologies and approaches in the age of Agentic AI and Large Language Models? There are many people who believe we can replace all of this “legacy” decision automation with AI. But the short answer is, you can’t. You still need the systems, but you can (and should) improve them with AI as there’s tremendous value to be unlocked.
Here are some examples of how we are helping our clients use AI safely:
- You can use AI to ingest and process unstructured data of all sorts: government documents, competitors’ forms, letters, information presented on forms you’ve never seen before – and in doing so, dramatically improve the time it takes to get the data you need. You are no longer limited to automating decisions within structured data sets.
- You can use AI to improve the presentation of your decisions too. You can generate natural language explanations and stipulations so that customers know what to do next and don’t need to call you to ask. You can make it easier to deliver consistency in customer service and improve the effectiveness of agents, investigators and adjudicators similarly.
- And increasingly you can use AI to help you define these systems, analyzing legacy code and policy documents to deliver first cut decision models ready for review and approval by you SMEs, or identifying how a document change impacts an existing decision model.
We are very excited about the future of AI and how Decision Automation plays a pivotable role in successful adoption and deployment. We have several exciting events coming up this summer, we hope to see you there!