AI In Financial Services: Three Current And Emerging Applications

  4 m, 50 s

While the impact of artificial intelligence (AI) is a bit of a mixed bag in a number of industries, we’re seeing some exciting traction in financial services. In this month’s article, I take a look at some specific examples of where machine learning and AI are helping financial services organizations improve their services, products, and processes.

AI Helps Financial Services Reduce Non-Disclosure Risk

Financial firms and banks are taking advantage of AI to ensure that their employees are meeting complex disclosure requirements.

Generally, financial advisors must make sure that their “client advice” documents include proper disclosures to demonstrate that they’re working in their client’s best interests. These disclosures may cover conflicts of interest, commission structure, cost of credit, own-product recommendations and more. For example, advisors must clearly disclose the fact that they’re encouraging a client to purchase a position in a company that the firm represents (a potential conflict of interest).

To ensure compliance, firm auditors randomly sample these documents and spot-check them by keyword or phrase searches. But this process is clunky and unreliable, and the cost of failure is high: Some estimates put the price of non-compliance as high as $39.22 million in lost revenue, business disruption, productivity loss and penalties.

To help financial services firms ensure disclosure compliance, companies like FINRA Technology, Quantiply and my company offer AI solutions that use semi-structured data parsing to analyze client advice documents and extract all of the component pieces of the document (including disclosures). Then, using natural language processing to understand the meaning of the underlying text, the AI structures this data into an easily-reviewable form (like an Excel document) where human auditors can quickly evaluate whether all necessary disclosures were made. Where before an auditor might spend hours to review 1% of their firm’s documents, AI solutions like this empower the same person to review more documents in less time.

AI Fights Elder Financial Exploitation

$1.7 billion. That’s the value of suspicious activities targeting the elderly, as reported by financial institutions in 2017 alone. In total, the United States Consumer Financial Protection Bureau (CFPB) says that older adults have lost $6 billion to exploitation since 2013. One-third of these people were aged 80 or older, some of whom lost more than $100,000.

Thankfully, tech companies and financial institutions are fighting back. The CFPB notes that “Regularly studying the trends, patterns and issues in EFE SARs [Elder Financial Exploitation Suspicious Activity Reports] can help stakeholders enhance protections through independent and collaborative work.” This is a great opportunity for machine learning and AI, which use reams of historical data to predict what is likely to happen next.

Wells Fargo, for example, uses machine learning and AI to identify suspicious transactions that merit further investigation. Ron Long, director of elder client initiatives for Wells Fargo Advisors, told American Banker earlier this year that their data scientists are constantly working to add new unstructured and structured data sources to improve their capabilities. “While a tool can’t replace human assessment,” he said, “machine-learning capabilities play an important part in our strategy to reduce the number of matters requiring a closer look so we can focus on actual cases of financial abuse.”

One example is EverSafe, an identity protection technology company founded in 2012, which draws on multiple data sources to train its AI. EverSafe places itself at the nexus of a user’s entire financial life, analyzing behavior across multiple accounts and financial advisors. This approach dramatically improves their AI’s ability to identify erratic activity or anomalous transactions. Eversafe’s founder, Howard Tischler, says he was inspired to create the company after his aging, legally blind mother was scammed multiple times, including by someone who sold her a deluxe auto club membership.

AI Adds A Crucial Competitive Edge In High-Frequency Trading

Back in the 1980s, Bloomberg built the first computer system for real-time financial trading. A decade later, computer-based high-frequency trading (HFT) had transformed professional investing. Some estimates put HFT at 1,000x faster than human-human trading. But since the 2010s, when trading speeds reached nanoseconds, industry leaders have been looking for a new competitive edge.

To keep up with (and ahead of) the competition, industry leaders are turning to algorithmic trading. The sheer volume of trading information available for machines to analyze makes artificial intelligence and machine learning formidable tools in financial marketplaces. Investment firms use AI to increase the predictive power of the neural networks that determine optimal portfolio allocation for different types of securities. In simpler terms: Data scientists use reams of historical prices to train computers to predict future price fluctuations.

AI has already proven its value in HFT. Renaissance Technologies, an early adopter of AI, boasted a return of 71.8% annually from 1994 to 2014 on its Medallion Fund (paywall). Domeyard, a hedge fund, uses machine learning to parse 300 million data points in the New York Stock Exchange, just in the opening hour. And PanAgora, a Boston-based quant fund, deployed a specialized NLP algorithm to quickly decipher the cyber-slang that Chinese investors use on social media to get around government censorship. These findings give PanAgora, a firm that operates at the speed of fiber optic cables, vital insights into investor sentiment fast enough to keep up with (and influence) its trading algorithms.

Wrapping Up: Tempering Expectations For AI In Financial Services

The value of AI in financial services is clear. But don’t get lost in the hype. For every useful AI system, you can find a dozen problematic algorithms and large-scale failures. To succeed, keep a realistic perspective of what AI can and can’t do to help.

The truth is that artificial intelligence is just a tool. Alone, AI doesn’t really “do” anything. What matters is how you combine AI with other technologies to solve a specific business problem.

This post originally appeared in Forbes Technology Council.

Categories: Lexalytics, Regulatory/Compliance, Special Interest, Text Analytics

Leave a Reply

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>