In the early stages of AI/ML, only the biggest and most innovative companies leveraged machine learning. Today, almost half the business world can be said to employ AI/ML as part of routine operations. According to Mckinsey’s “The State of AI in 2021” survey, 56% of companies surveyed had adopted AI - claiming to represent “the full range of regions, industries, company sizes, functional specialties, and tenures.” In 2022 and beyond, this percentage only continues to increase.
AI adoption is even more advanced in the financial industry, for both traditional financial services firms and newer FinTech companies. According to a study from the World Economic Forum and the University of Cambridge Judge Business School in 2019, 85% of survey respondents had adopted some form of AI.
As AI and ML become increasingly prevalent, how can we forecast future trajectories in the FinTech market? We’re expecting to observe 3 broad trends to continue to develop in the coming months.
- AI will make increasingly critical decisions.
- AI will democratize access to financial services.
- Companies that can consistently deploy high quality AI models will win in the market.
Let’s unpack some of these movements.
1. AI WILL MAKE INCREASINGLY CRITICAL DECISIONS
Within the business world, AI in business was first used for algorithmic recommendation systems - suggesting certain products based on previous google searches, for example, or promoting ‘relevant’ accounts on social media platforms.
As artificial intelligence has developed further and become better adjusted to business processes, however, more complex tasks have been outsourced to AI and ML capabilities. In 2022, AI is used in the FinTech space to help make automated, mission-critical decisions. Applications within issuing credit, pricing insurance and loans, and detecting and blocking fraudulent transactions are becoming the norm.
Leading companies in the space are leveraging AI as a way of reducing the bottleneck-effect of human decisionmaking. Yet, giving more power and responsibility to AI comes with a host of implications, ranging from ethical decision-making, to pending regulatory measures, to shaping essential private and public services.
Public perception is hyper-sensitive to failures of AI. Any suggestion of biased decisions or unethical use of data can turn into a PR nightmare for companies, and can have very real negative impacts on end-users. When it was first launched, the Apple Card was subject to allegations that it was issuing up to 20x the credit limit to men than it was to women. This quickly made it to the front page of the Wall Street Journal and countless other news outlets and blogs.
Lemonade, an AI-powered insurance company, found itself in hot water after a Twitter thread was accused of suggesting that the company was using people’s appearances to judge whether or not their insurance claims were fraudulent. Given the company uses its bot “AI Jim” to handle one-third of insurance claims from start to finish, any suggestion that the company was taking physical appearance into account when making these decisions very understandably caught the attention of the public and the press.
AI is poised to make increasingly important decisions within the FinTech universe. Along this momentum, we have to recognize that trust in these systems is still brittle; it is imperative that we implement robust processes alongside increased automation networks, ensuring that they are trustworthy enough to continue operating as intended once they are launched into production systems.
2. AI WILL DEMOCRATIZE ACCESS TO FINANCIAL SERVICES
In the coming months and years, AI in FinTech will continue to democratize access to key financial services for historically underserved markets.
Take the hypothetical example of a young adult who has no credit score in the USA - perhaps they recently immigrated, or their parents simply never helped build a credit score for them. Historically, they may have faced an arduous journey to build access to credit, often starting to build a credit score via secured cards to eliminate risk for the lending institution, before getting an actual credit card. This was a slow process that hindered their ability to rent a place to live and gain further access to credit.
In 2022, FinTech companies have begun to use machine learning and data science to subvert traditional barriers to entry for their customers. The credit card provider “Petal,” for example, has recently become successful in its efforts to leverage machine learning as a way of more efficiently analyzing digital financial records. Instead of relying on a person’s credit score as the only relevant criterion, Petal scrutinizes the personal cash flow of applicants, and uses machine learning to determine that person’s risk score and issue an appropriate level of credit.
Other examples in the market include Wealthfront, an automated investment service firm based in California, which has begun using AI and ML as part of their Data Analytics Platform for automated portfolio management processes. In the past, only individuals who could reasonably expect to pay for their own investment manager would expect routine overseeing of investments. Now, AI is doing all of it for us - allowing a greater number of people to maximize their portfolio value without additional investment.
Similarly, the Brazilian neobank giant Nubank is becoming a leader in Latin America for using ML to evaluate how much credit to give their customers. Across the world, more and more FinTech companies large and small are adopting AI and ML as ways of improving their system operation and democratizing their services.
3. COMPANIES WITH HIGH QUALITY AI MODELS WILL WIN IN THE MARKET
The task of building, deploying, and operating machine learning and artificial intelligence systems has become more streamlined over time. Today, the major hurdle that most systems face is learning how to deploy ML models in production in a method that ensures performance consistency.
Performance and trust go hand in hand. ML systems create automated decision-making processes, which can incorporate and reinforce biases and fall foul of issues like data drift or broken data pipelines. When performance is compromised, ML systems lose trust from both internal non-technical stakeholders, and with customers.
In order to maintain this trust and optimize model outcomes, companies cannot compromise on the quality of the testing and monitoring processes when deploying models into production. After all, issues in production can result in very real impacts for customers and the company’s bottom line.
In November of 2021, for example, the real estate firm Zillow was forced to exit the home flipping business as a result of failures in their algorithmic models for buying/selling homes due to shifts in real world economic conditions driven by COVID-19. For FinTech companies like Zillow to succeed, they need to do everything in their power to eliminate any failures of their AI models.
RI, AI, and the Future of FinTech
AI will continue to make increasingly critical decisions, democratize access to financial services, and set apart best-in-class technical companies from their competitors. For those fintech companies that are unable to effectively leverage AI, the exact opposite will be true. They will be stuck trying to build basic models, failing to get their models into production. Ultimately, they will fail to drive business impact through machine learning, and will fall further behind their more technically advanced competitors.
This is where Robust Intelligence can help.
The Robust Intelligence Model Engine and AI Firewall instill best practices into the model development and deployment process. You can automate testing and validation of models pre-production, and monitoring and active protection of models once in production, ensuring they are operating as intended, and flagging when retraining is needed (request a demo here!).
As ML continues to dominate the fintech space, winning players in this market will be the ones who can reliably leverage ML to drive business impact. By automatically preventing AI Failures using RIME, machine learning teams can focus on building models which move the needle for their business.