By David Sosna
There’s growing interest in AI adoption across industries, and banking is no exception. Many in the field are testing the waters, seeking ways to put the technology to effective and productive work.
But banking has unique characteristics that can challenge efforts to gain value if not hard ROI from these investments — principally its highly regulated environment and correspondingly risk-sensitive nature.
Like the Hippocratic Oath, the immediate priority for banks seeking to use the latest AI innovations should be: “First, do no harm.”
A brief history of AI in banking
AI in banking is not a new phenomenon. Financial institutions have relied on artificial intelligence for decades, initially through rules-based expert systems for credit scoring, fraud detection and risk management in the 1980s and 1990s. The rise of machine learning, big data and cloud computing in the 2010s spawned increasingly sophisticated models to power personalized recommendations, chatbots and algorithmic trading.
The introduction of ChatGPT in late 2022 brought generative AI into public consciousness, accelerating efforts to tap into generative AI. Financial institutions have been moving beyond pilot programs to implement these advancements at scale, in customer service, risk management, and operational efficiency.
But the road to success can be littered with failed efforts. Increasing sophistication and automation carry increased risks. Also, common myths and misperceptions might confuse about what to expect and misdirect investment and efforts.
As a serial entrepreneur and CEO who has built companies that applied data and analytics to solve problems, I have seen tech and trends come and go. I have been closely watching the rapid advancements in GenAI and agentic systems, and the potential of his technology to transform banking truly intrigues me.
To reap the rewards, it is important to counter myths with a more realistic sense of what is achievable. I thought I would share a few that we keep running into, with the hope that a better understanding will lead to better results.
Myth 1: AI mostly helps in cutting costs
Many think that AI is primarily for cost-cutting, whether that means reducing headcount (the elephant in the room, rarely said aloud), or increasing back-office efficiency. In reality, the most interesting applications emerging are about growing revenue. For example, AI can be used to personalize financial products, predict customer behavior and optimize marketing, which can boost sales and customer loyalty, according to this CMSwire piece.
Another option is to scale relationship banking with AI. GenAI and agentic architecture can help banks to scale efficiently and cost-effectively by automating routine tasks, uncovering client needs, and equipping bankers with the right insight and guidance for advising, cross-selling and upselling clients.
There’s a big opportunity here, considering that small and midsize business banking generates about $150 billion annually, or roughly 17% of the overall U.S. banking industry’s revenue. There are more than 30 million U.S. SMBs. A McKinsey study concludes that SMBs are increasingly seeking a blend of robust digital tools and strong relationship management when choosing their primary bank.
Ultimately, it’s not about replacing staff, as there just aren’t enough of them; it’s about helping them be more efficient and scaling their efforts to deliver quality advice.
Myth 2: AI replaces banker insight and knowledge
Some think that gains for AI will mean losses for people (yes, that jobs concerns again); and that insight provided by tools like GenAI can compete with and erode the need for human insight.
In reality, it will likely be some time (if ever) before AI can replace the high-touch service and personalized communication that builds trust and fosters relationship-building. Sure, it can empower bankers with critical info and increase their efficiency. AI can be used to scale relationship banking and helps address the talent gap for this role.
It can do this by helping users access information and glean useful insight, from internal and external data sources. The models learn over time and can provide guidance based on institutional knowledge of relationship management best practices.
Of SMBs surveyed, 47% said that dedicated relationship manager support is a key decision-making factor. AI can scale this support and help banks better meet the needs of these customers.
Myth 3: It’s all about GenAI
As the saying goes, when you have a hammer, everything looks like a nail. Here, the hammer is GenAI. Its most famous version, ChatGPT, lit a fire under AI adoption and fueled the innovation race that brings us to today.
One could be forgiven for thinking that GenAI can be applied everywhere to solve all related AI problems. It’s your consumer-friendly AI assistant that seems to get smarter and more versatile every day.
While GenAI may be the low hanging fruit, banks should take a hard look at other kinds of AI solutions and use them where appropriate for the best results (to say nothing about avoiding the risks that GenAI can introduce, in the form of hallucinations and security holes)
Here are some examples:
- Agentic AI systems autonomously plan, decide and act toward a goal. They are increasingly replacing or integrating with earlier robotic process automation technology. Unlike RPA, which follows fixed rules and scripts, agentic AI employs multi-step reasoning and can interact with systems and data sources in real time. In banking, it can automate relationship management tasks, assist in responding to customer queries, and orchestrate workflows (say, for onboarding or compliance resolution).
- Machine learning discerns patterns from historical data to make predictions or decisions, for example in credit scoring and fraud detection.
- Predictive analytics uses statistical models and machine learning to forecast future outcomes, for product recommendations and cash-flow forecasting,
- Anomaly detection identifies unusual patterns for real-time fraud monitoring.
GenAI is just one form of the technology; there are other types of AI that provide substantive results.
David Sosna is a serial entrepreneur with over 20 years of experience building innovative fintech startups, including Actimize, Personetics and Sympera AI.
