AI In Finance Is Here; Is It Working?

Reco McCambry, CEO @ Novae, a black-owned fintech providing greater access to credit, capital & entrepreneurship for underserved communities

2025 has been the year of AI. While genuinely intelligent machines aren’t here yet, programs designed to analyze huge amounts of data are being rapidly deployed in every industry.

But how beneficial is today’s AI in financial applications? Many reports about the results of applying AI are written by companies that sell AI, and independent researchers are still catching up. To supplement the sparse independent analyses, I spoke to members of my fintech team and to bankers who use AI for work.

Their findings? AI does offer unique value adds that far exceed human capabilities in some applications, but in other applications, it actually appears inferior to human workers.

In one survey, 96% of executives said they expected AI to improve their business’s productivity—but 77% of employees said AI tools slowed them down. AI was beneficial when used to do tasks humans couldn’t perform, such as real-time analysis of huge amounts of data and taking care of tasks that were completely unstaffed before. But introducing AI into simpler workflows such as customer service and communications introduced errors and actually slowed things down.

So what is AI doing in finance, and how well is it working?

Fraud Detection

One use for AI in finance is fraud detection. To fight new methods of identity theft, increasingly aggressive fraud detection algorithms have led to increases in erroneous fraud reports. These mistaken reports can result in accounts being frozen and payments being canceled at crucial moments.

Is AI working to improve the situation? The answer appears to be “yes,” assuming the AI program is well-constructed. For example, according to an analysis by risk management and business strategy firm OliverWyman, PayPal has been able to cut incorrect fraud reports in half by replacing older fraud detection algorithms with AI.

Algorithmic And High-Frequency Trading

Another growing use of AI is in automated stock trading. As with fraud detection, trading had already been automated using algorithms before AI. So are trading AIs performing better than the old algorithms?

Again, the answer appears to be “yes.” According to several studies and literature reviews, some types of AIs have increased the accuracy of market prediction and created better investment returns. However, the authors of the study caution that AI doesn’t have the human ability to understand how current events may affect market behavior. Because of that, it’s wise to ensure that trading AIs are subject to oversight by humans with the ability to pause and override decisions.

Personalized Banking Experiences

From customer service chatbots to personalized recommendations, AI has already changed most people’s banking experiences. So how do customers feel about this, and how is it affecting banks’ bottom lines?

Surveys show that most customers dislike being routed to a chatbot instead of a human when they need help. In one survey, only 12% of customers preferred AI chatbots over human customer service agents, while 48% preferred live humans in all circumstances.

However, customers do like receiving personalized offers from their banks. A survey by data analysis firm MX found that 54% of customers wanted banks to use their personal data to offer rewards and products optimized for their financial situations.

So does AI succeed at delivering better personalized experiences than older methods?

McKinsey’s State of AI 2025 report states that companies using AI to personalize recommendations saw improvements in lead identification, customer journey mapping and other key steps of the sales process. This makes sense; AI can collect and analyze more data than older algorithms, so it should produce better recommendations if everything is working right.

But be aware of potential conflicts of interest in these reports. McKinsey profits from selling AI to businesses, and from my observations, their report seems to be the only source cited by dozens of articles about AI improving sales outcomes that have been published across the internet in 2025.

Error Detection And Correction

One use case for AI in finance that is still being explored is utilizing AI agents to assist customers with error remediation. According to a study involving 4,300 volunteers, almost half of Americans’ credit reports contain at least one error, and that number has been rising rapidly for years. Back in 2012, the proportion of credit reports with errors was “only” 20%.

Some errors are relatively harmless, but others can result in being denied financing, experiencing inflated interest rates or being denied employment. However, even items that are negative but accurate can often be removed from one’s credit report by negotiating a payment plan with the creditor. Many Americans don’t realize they have that option, or don’t bother because it seems too hard.

My own company has seen early success in using AI agents that work for customers to resolve such errors, even reaching out to credit bureaus on their behalf. This does require that the AI be trained in escalation techniques and legal language that are often necessary to obtain a credit report correction. From what I can see, offering chatbots that work for customers in this way is not yet widespread in the finance industry, but I believe that could change in the future.

Conclusion

Early results from the use of AI in finance look promising. AI tools appear to be more effective than older algorithms at analyzing large amounts of data in real time and making accurate predictions about market and consumer behavior. These are tasks that humans arguably can’t perform effectively, making these AIs a unique value add.

However, many customers and workers report that inserting AI into tasks humans can perform can make matters worse. Workers report that inserting AI into established workflows can actually decrease productivity, while customers find that AI chatbots provide inferior customer service experiences compared to human agents.

I believe the lesson we should take away from this is that companies should be judicious in how they employ AI. Take advantage of its unique abilities, but don’t alter your human-centered workflows if they aren’t broken.

Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?