The use of AI is increasingly relevant in innovating banking and finance, an area producing loads of underutilized, unstructured data. This data could hold valuable insights to assess risk and even guide investment decisions. New no-code AI automation capabilities could open these insights more readily, leveraging pre-trained models to democratize AI and bring significant innovations to the finance world. Similarly, sector-specific ML models can empower knowledge workers supporting business units across many other industries.
I recently met with Anshul V. Pandey, Accern co-founder, and CTO, to discuss how the world of banking and finance is being reshaped by new AI automation. According to Pandey, applying deep learning within the finance industry has become a game-changer for mining valuable insights from unstructured data. But, to truly accelerate financial services, AI automation will require ease of use via no-code. Also, it will require customization options so that financial services can leverage the “secret sauce” that gives their business a competitive advantage.
Industry-Specific No-Code AI Automation
Historically, Natural Language Processing (NLP) models have been quite difficult to train. One reason is that most companies don’t have nicely assembled data for a training model to easily ingest. It’s estimated that 80–90% of corporate data is unstructured. Unstructured data encompasses text, images, video, audio, or other files that don’t possess a standard data model.
Deep learning models have become a game-changer for utilizing unstructured data, Pandey notes. However, “general-purpose models don’t scale well across business types,” he said. For example, training a model on healthcare user data won’t apply well to finance user data. Similarly, training a model to detect sentiment from eCommerce product reviews won’t translate to a healthcare setting.
Therefore, Pandey advocates using pre-trained sector-specific ML models familiar with unstructured data unique to that field. By honing in on the unique characteristics of data from a particular industry, ML models can be more tuned to the situation at hand. Furthermore, as training these models is computationally taxing and require skill to develop, a no-code approach could improve usability. Applying low-code/no-code capabilities to automate AI analysis would help democratize access to AI, helping business units process and respond to data in real-time.
Benefits of Using AI for Innovating Financial Services
“People who are riding this AI bandwagon are definitely seeing an edge,” said Pandey. So, what sort of benefits could no-code AI automation bring to the financial sector? Well, applying AI to a financial setting could help gauge risk to guide better investments.
For example, Pandey describes how AI aids underwriters in assessing a loan application. By analyzing unstructured data associated with each case, like customer service notes or transaction histories, underwriters could spot anomalies. For example, knowing that a large spending was for an ongoing lawsuit, versus a present to a family member, could carry stark implications for the loan approval process. With more informed insights, lenders could reduce loan default rates.
Similarly, AI could be applied to other credit approval scenarios, such as bank accounts or credit cards. For banks, AI could be a big boon to better profile potential customers. But financial services are also utilizing AI to inform large-scope investment decisions too. For example, by mining unstructured public data for sentiment, a hedge fund can gauge collective attitudes around specific companies, brands, or trends and invest accordingly.
Tips For Leveraging AI in Innovating Finance
Use pre-trained industry-specific models. While there are open-source libraries to train ML models in-house, Pandey stresses there is a competitive advantage of using pre-trained models. R&D departments within financial institutions must constantly balance two pressure points: keeping costs low and quickly demonstrating a proof of concept for new ideas. Utilizing pre-trained models and ready-made SaaS could help avoid reinventing the wheel and keep costs down for introducing new AI-driven innovations.
Adopt no-code to provide AI capabilities directly to end business units. There is a strong correlation between the proximity of the project to the end business unit, and the use of the generated insights to drive positive business outcomes, notes Pandey. Therefore, no-code-style AI solutions should sit close to end business units to best serve knowledge workers.
Ensure customization to retain secret sauce. “In finance, every firm has their own secret sauce,” said Pandey. “They wouldn’t have an edge without it.” Many firms have already done much internal research around AI uses cases. Some even possess tagged information or classes they want to utilize within their own models. To retain unique flavors, ensure the AI automation tools you use provide customization options that go beyond generic procedures. The ability to customize automatically cuts down the need to build your own data engineering pipelines, Pandey described.
Ensure side doors for programmability. While a no-code approach is great for improving usability for non-engineers, you don’t want to be permanently stuck with UIs and vendor limitations. Comprehensive low-code/no-code platforms also enable deeper machine-machine programmability to enable extensibility with external stacks. “Going under the hood it’s very important,” said Pandey. For example, Accern accomplishes this by allowing pro-coders access to a backend engine framework.
AI to Empower More Business Units
Software development is constantly traversing up the stack with more levels of obfuscation. Similar to how you no longer need to write in an assembly-level language to generate applications nowadays, similar obfuscation is occurring in the realm of AI. Now, through a low-code/no-code approach, AI power can reach the hands of citizen developers in addition to data scientists. For banking and finance, the inclusion of AI could lead to more intelligent background checks and funding decisions, thus decreasing risk and ensuring higher returns on investment.
“We are in the early adoption stage of AI-powered automation,” said Pandey. Looking to the future, no-code will likely contribute to accelerating the democratization of AI. Pandey also foresees similar adoption of AI to power business units across other sectors. “No-code AI will become the defacto.”
