AI progress and adoption are accelerating – four factors banks should consider

Artificial intelligence (AI) is progressing quickly, with new developments reshaping industries on a global scale. Recently, DeepSeek, a Chinese AI startup, unveiled its R1 model and disrupted the AI landscape. Meanwhile, Manus, another China-based AI creator, is gaining increased attention with an advanced AI agent capable of handling complicated tasks such as e-commerce data analysis, supplier sourcing and stock market research.

From the standpoint of financial services, and specifically, banking, these latest competitors appear either largely on par with U.S. counterparts to date or still in limited beta and unavailable for comprehensive benchmarking. But they do represent something very important: a larger trend, and more evidence that AI progress and adoption are rapidly accelerating.

As more of these solutions begin entering the marketplace, banks must carefully assess the impact and know how to incorporate AI into their operations. Here are four key considerations for banks and their leadership regarding current AI trends.

1) A case for self-hosted AI models

AI technology is becoming more accessible, giving banks the opportunity to deploy self-hosted models for enhanced security and compliance. Typically, self-hosting is expensive, particularly for leading models, with hardware expenses exceeding $100,000. However, trending cost reductions of 1.5x to 4x year over year could make self-hosting AI an economically viable option for banks.

Most banks rely on AI solutions from major tech firms that offer contractual privacy protections but do not provide physical control over data security. Self-hosted AI is an attractive alternative for banks with strict regulatory requirements, as this option gives complete oversight over sensitive customer data. For banking leadership, the decision to self-host AI could help mitigate privacy concerns and gain an edge over their competition.

2) AI’s expanding role in fraud prevention and customer experience

AI is already making substantial improvements in baking security and customer service. For instance, South State Bank implemented Tate, an AI-powered knowledge management bot that’s reducing employee search times from seven minutes to under 30 seconds. The Commonwealth Bank of Australia is also leveraging AI to scan over 20 million daily transactions, helping detect fraudulent activity and reducing customer wait time by 40%.

AI’s ability to analyze vast quantities of transaction data in real time is crucial for mitigating risk, especially with fraud schemes becoming more sophisticated. Moreover, AI-powered customer service tools streamline banking interactions, offering personal assistance and enhancing overall satisfaction. Banks that fail to proactively adopt AI-driven fraud prevention solutions risk falling behind in a hyper-competitive market.

3) Address internal and external risks

Internal risks from AI include misplaced trust in generative AI’s accuracy and bias, along with the misuse of unsecured tools for sensitive data. Employees are using AI to automate parts of their work and enhance efficiency without their institution’s approval. There are serious risks for banks that don’t implement a strategy and policy to foster safe adoption with appropriate oversight.

In terms of external risks, malicious actors are leveraging AI for similar efficiency advantages that result in increased levels of phishing, fraud and cyberattack volume, personalization, and sophistication. All these factors will impact FIs whether they choose to adopt internally or not.

4) Balance AI investment with rapid technological change

For small banks, taking a “wait and see” approach is a viable strategy for certain types of solutions and use cases. However, there are areas such as training, development, marketing and strategy in which banks should act quickly.

There’s no denying AI’s transformative potential, but banks must be strategic in their investment to avoid obsolescence. AI models evolve rapidly, resulting in cutting-edge solutions today becoming outdated within a year. Bloomberg experienced this firsthand after investing early in a custom large language model (LLM) designed for financial queries. This LLM was quickly surpassed by OpenAI’s general-purpose GPT-4.

Banks must focus on adaptable AI solutions rather than locking into rigid proprietary models. Prioritizing AI adoption in areas like risk management, marketing, and automation enables banks to stay ahead while maintaining flexibility to integrate future advancements.

Stay prepared for AI’s continued evolutions

AI is no longer a distant innovation – it is actively shaping the future of banking. While some applications require careful evaluation, banks should actively integrate AI solutions that enhance efficiency, security and customer satisfaction.

Institutions that proactively embrace AI will be well-positioned to capitalize on the next wave of advancements, ensuring they remain competitive in an increasingly digital financial landscape.

Connor Heaton is the Director of Artificial Intelligence at SRM.