There’s no denying it anymore: today’s global economy is going digital. In the past five years, FinTech investments have drastically increased across industries, and banks in particular are taking interest in this cutting-edge technology. According to the 2018 Digital Banking Report, more than half of banks reported big data, artificial intelligence, advanced analytics and cognitive computing as their top priority moving forward.

These priorities are shaping the banking industry at this very moment. In fact, 30 percent of large financial institutions had already invested in AI as of 2017 — and that number is expected to rise in the coming years. But for developers of financial products and services, there’s a growing list of questions to consider when creating and adopting FinTech, including:

• Does it meet banking regulations and compliance requirements? If not, what compliance services should I turn to?
• How do I verify user identity?
• How long will it take to bring in bank partnerships?
• Which manual functions could be streamlined with machine learning?
• How can I mitigate my risk, as well as my perceived risk for partners?

The list goes on. Let’s dig into how machine learning answers some of these questions — as well as some of the roadblocks banks face in adopting machine learning, and the future of AI as it stands today.

Top 3 Challenges for FinTech Companies

FinTech companies attempting to provide financial products and services will inevitably face a variety of obstacles. However, there are three issues that consistently come up: bank partnership, compliance and integration.

1. Bank partnership

Any FinTech company looking to go to market with some kind of finance product, such as savings or investment applications, will need to partner with a bank. But because of the many technical pieces and regulatory issues it involves, bank partnership discovery is a long and arduous process; if you attempt it by yourself, it can take anywhere from six months to a year.

2. Compliance

The engineers building banking platforms for FinTech companies are highly technical but don’t understand the banking landscape quite as well. The regulatory landscape, in particular, is vast, dynamic and varies from region to region and industry to industry.

While most FinTech companies have a great amount of expertise on the technical side, they often don’t understand how to implement their products in compliance with governmental and other regulatory agencies — making banking partnerships that much more difficult.

3. Integration

Last but not least, one of the biggest challenges for FinTech companies is integration. As technology becomes increasingly complex, both businesses and consumers are looking for the one-stop shop for all of their needs.

They want the enterprise solution that eliminates the need to go to market with multiple partners. But this requires fairly robust control over FinTech technologies, and it adds another layer of complexity to banking compliance.

Banks are Lagging Behind in Machine Learning

All of these challenges can be addressed with AI or machine learning tools. In today’s fast-paced, technology-driven world, it seems like these are obvious issues that traditional banks should be equipped to handle. But they’re lagging so far behind in the field of AI that many customers still have to physically walk into a bank with a stack of paperwork to provide verification.

Actually, this isn’t an unusual issue. Because banks have to comply with various governmental agencies and other regulatory entities, automating identity verification and other manual processes with AI is not on the top of their to-do list. In fact, 54% of financial institutions named data storage, privacy, and protection as their main barrier to innovation. Not only that, but most of these regulations are fairly new to banks. In the U.S. especially, we may make the mistake of thinking of these as long-standing protocols, but many of them were developed after 9/11.

Furthermore, the bank executives making decisions on these software solutions tend not to be technically-minded themselves. This is made worse by the fact that machine learning is a fairly new technology; in fact, most of these technologies didn’t exist until just a few years ago. In the case of computer vision for identity verification and due diligence, it just wasn’t as good two years ago as it is now. If you’re not necessarily a software guru, like a bank executive, you might not have known.

If that weren’t enough, budgetary issues often get in the way as well. Budgets are carved out year by year for exploring new products or projects, but the money isn’t always utilized in the best way, and banks aren’t allocating appropriate funds to implement automation.

FinTech Gaining Speed: Machine Learning for Individualized Decisions

With banks lagging behind on these issues, FinTech companies are innovating for speed, ease and accuracy.

There are competing schools of thought on best practices. Some, including all of the credit score companies, still really believe in one centralized identity verification solution — but these companies have a lot of false negatives and bad data that could be cleaned up with machine learning and a risk-based probabilistic approach. After all, SSNs have already been compromised for half the country.

Others think that centralized systems have been rendered unnecessary by machine learning. The whole reason that we had a centralized identity verification system was because we use a number, such as a credit score, that ten other companies have relied upon to make an evaluation of a potential customer’s situation. It was once used as a proxy for the inability to recreate manual work again and again by humans, but now the cost of computation with AI is so low, we don’t need to rely on these legacy databases — we actually can rewrite everything every time.

Let’s take account creation, for instance. Instead of someone having to physically walk into a branch with a utility bill, banks could now use machine learning and computer vision to inspect these documents online and make automated, individualized decisions.

Years ago, this wasn’t a possibility. But now? AI can take the documents, complete the underwriting and create an account without the individual ever having to set foot into a bank. Machines can automatically do the old school manual work of facial recognition, lip reading and real-time income verification — and they can perform these tasks again and again without the risk of human error. With another layer of computer vision and automation, we may even enhance due diligence.

What’s Next?

There’s been a lot of push on the consumer side for FinTech solutions, the most common and popular of which is Venmo. Consumer FinTech solutions like these have been lauded for their ease of use, intuitive design, faster service and increased accessibility. They’re not going away, either; 72 percent of banks reported enhancing the digital experience for consumers was their top priority for 2018. So why isn’t there a B2B version of Venmo?

In short, the answer is that business verification and transactions are much more complicated and would require more advanced computer vision. To verify an individual consumer, one could use machine learning to scan their government ID against their selfie and their address against the location tracking services on their phone, for example. But approvals for businesses require articles of incorporation, bylaws, identification of all the primary shareholders and more. These are sometimes fifty-page documents, as opposed to a consumer’s one-page utility bill. That means we’ll need more creative solutions.

In the coming years, more and more manual processes will be automated for faster, individualized and non-traditional service. The bottom line is that FinTech is evolving every day — and the banking industry is evolving with it.