- Most banks are still experimenting with generative and agentic AI, leaving them vulnerable as the technology erodes traditional advantages like customer stickiness.
- AI is shifting power to consumers by increasing transparency, improving underwriting and challenging banks’ advisory services.
- To stay competitive, banks must integrate AI into their core strategy, upgrade tech and governance, and upskill their workforce for AI-driven transformation.
Banks are facing a critical juncture as generative and agentic AI chip away at their traditional advantages, according to a new BCG study. But rather than embracing the technology strategically to reshape their businesses, most banks are merely experimenting at the edges.
“Oftentimes, it’s not fully integrated into a business strategy,” said Stiene Riemer, managing director and partner at BCG, in an interview with The AI Innovator. “Sometimes, it’s not fully led from the top.”
Only a quarter of banks surveyed are using AI to bolster their competitive position, with the rest still in pilots or proofs of concept.
While financial institutions have been using predictive AI for decades, generative and agentic AI – the recent entrants – are dismantling banks’ traditional moats. Banks have historically benefited from complexity.
“Customers stayed put, pricing structures weren’t always clear, and financial products were tied to proprietary distribution channels,” Riemer and her co-authors wrote.
But generative and agentic AI will shatter this opacity – and give banks a run for their money. For example, the report’s authors observed the following:
- AI-powered agents will make it easier for customers to find better deals at competitors. “Banks that once relied on stickiness will need new ways to earn loyalty,” the authors said.
- AI will drive transparency in other areas, revealing rate structures, fees and loan terms in real time.
- AI-powered financial decision-making shifts the control back to the consumer, making it harder for banks to own the customer relationship.
AI will make underwriting and credit risk assessment more transparent and timely, potentially reducing the margins banks can charge on loans. Banks advisory services in financial planning also would be affected if they cannot do better than the AI assistant.
Banks also face technical barriers such as legacy infrastructure, data fragmentation, and concerns about AI model reliability. The complexity and potential risks associated with generative models – which can produce ‘hallucinations’ or inaccurate outputs – have prompted caution.
However, Riemer emphasized that banks are uniquely positioned to manage these risks due to their strong existing frameworks for model governance and risk management.
“Banks, from the start, have a risk function. … They have a very well set-up model risk management framework,” Riemer noted, highlighting their advantage in proactively shaping governance policies. The key, she argued, is actively engaging with regulators to establish clear standards, a step many institutions have yet to fully embrace.
BCG’s report identifies crucial steps for financial institutions seeking to harness AI’s potential:
- Upgrade your strategy: Figure out where AI can give you a defendable advantage.
- Put AI at the center of tech and data: Invest in foundations to unlock the ROI of AI.
- Own the governance agenda: Create risk management frameworks and engage regulators proactively.
- Realign talent and accountability: Plan for coming organizational changes due to AI.
As for hiring workers skilled in AI, Riemer explained that while hiring specialized AI developers remains a challenge, banks should not neglect upskilling existing employees to interact effectively with AI systems. “What we see is that in a large number of institutions, only a very small percent of their workforce, even if they are consumers of AI, have been fully trained in the usage,” she said.
Riemer pointed to three particularly promising use cases for banks leveraging generative AI: enhancing customer service with conversational AI, accelerating software development cycles, and transforming cumbersome processes like Know Your Customer (KYC) checks.
What’s coming is “conversational banking, and even moving further into agentic selling, and having fully automated interactions with customers to provide even better and targeted services,” she stated.
Despite the current hesitation by banks, Riemer highlighted that successful deployments of agentic AI, particularly in wealth management, already exist. For example, some institutions employ multi-agent systems to support relationship managers through automated product recommendations, risk assessments, and client proposals.
Ultimately, banks must move urgently and decisively. She said institutions must prioritize strategic transformation over isolated technology deployments. Identifying key business outcomes, prioritizing a manageable number of use cases, and establishing clear KPIs are pathways to success.
