How the Agentic AI Revolution is Transforming Operations at 70% of Banks
By David Evans, Chief Content Officer at The Financial Brand
Simple Subscribe
Subscribe Now!
Executive Summary
- Rapid adoption is already underway: 70% of banking institutions are using agentic AI through existing deployments (16%) or active pilot projects (52%).
- Fraud detection and security lead use cases: More than half of executives report high capability in fraud detection (56%) and security (51%), with banks using AI agents to continuously monitor suspicious activities and automatically respond to threats.
- Human oversight remains essential: While 95% say AI systems can advise and 92% can assist, only 38% believe current technology is capable of full digital autonomy, positioning agentic AI as an advanced assistant rather than replacement.
- Implementation requires incremental approach: Leading institutions start with simpler agentic systems addressing specific process needs, then scale over time to automate larger workflow components, avoiding the complexity trap that can derail ambitious projects.
According to a new 2025 MIT Technology Review survey of 250 banking executives conducted with EY, banking is experiencing a fundamental shift as agentic AI moves from experimental technology to operational reality. Unlike previous AI implementations that required constant human guidance, agentic AI systems can analyze complex data patterns, make contextual decisions, and execute actions autonomously — transforming how financial institutions operate and serve customers.
The Technology That Changes Everything
Agentic AI represents a quantum leap beyond traditional automation tools like robotic process automation. These systems can process vast quantities of unstructured data, understand context, and make decisions that previously required human intervention. As Sameer Gupta, Americas financial services AI leader at EY, explains: “With the maturing of agentic AI, it is becoming a lot more technologically possible for large-scale process automation that was not possible with rules-based approaches before. That moves the needle in terms of cost, efficiency, and customer experience impact.”
The applications span the entire banking ecosystem. From responding to customer service requests and automating loan approvals to adjusting bill payments based on paycheck schedules and extracting key terms from financial agreements, agentic AI is reshaping both customer experience and internal operations.
Current Applications Delivering Results
Banks are already seeing measurable success across multiple functional areas. Fraud detection leads the pack, with 56% of executives reporting high capability in this area. The technology continuously monitors employee and customer behavior, identifies potential bad actors using sophisticated pattern recognition, and can even detect AI-powered fraud attempts like deepfakes.
Security applications follow closely, with 51% reporting high capability. These systems automatically respond to anomalies by quarantining infected machines, manage identity verification processes, and provide real-time threat detection that scales beyond human monitoring capabilities.
Customer experience improvements, cited by 41% as highly capable, include intelligent customer service routing, personalized financial advice, and streamlined onboarding processes. DBS Bank, for example, uses agentic AI to synthesize and classify complex SWIFT messages containing financial transaction instructions, presenting actionable information to humans for approval.
The Intelligent Assistant Model
Despite the technology’s sophisticated capabilities, banking leaders emphasize a collaborative rather than replacement approach. Ian Glasner, group head of emerging technology at HSBC, describes the optimal relationship: “Think of agentic AI as like an intern helping you get all of the more simplistic tasks done, but the human is still there to oversee and take the final decision.”
This philosophy aligns with current capability assessments. While 95% of executives say their AI systems can effectively advise and 92% can assist, only 38% consider them capable of full digital autonomy. The technology excels at processing information, identifying patterns, and preparing recommendations, but human judgment remains crucial for final decisions, especially in high-stakes situations.
Overcoming Implementation Challenges
The path to successful agentic AI adoption involves navigating several critical obstacles that can derail even well-funded initiatives.
The governance challenge: Managing governance, risk, and compliance represents the single biggest challenge for 63% of banking executives. As regulations lag behind rapidly evolving AI capabilities, institutions must implement their own guardrails to ensure responsible deployment.
Leading banks are developing comprehensive frameworks to address this challenge. HSBC maintains detailed inventories of AI systems linked to business owners, model documentation, and risk classifications. DBS Bank operates under a “PURE” framework requiring all AI systems to be Purposeful, Unsurprising, Respectful, and Easy to explain.
For high-risk applications, DBS implements real-time metrics with defined performance limits. If any metric is breached, automated kill switches activate, ensuring human oversight remains paramount in critical situations.
The skills and cultural adaptation gap: The second most common challenge—lack of technology skills and capabilities—affects 58% of institutions. This extends beyond technical training to fundamental organizational change management. Banks must simultaneously educate their workforce on AI capabilities while overcoming resistance to new working methods.
Successful implementation requires what Gupta calls improving “the general IQ and EQ of the organization” to adopt AI systems effectively. This involves training existing staff to work alongside AI agents, measuring ROI convincingly to overcome C-suite skepticism, and creating feedback loops that demonstrate value.
Data integration and quality issues: Poor data quality and integration challenges affect 54% of institutions, reflecting the complexity of modern banking infrastructure. Large financial institutions operate hundreds of existing data systems requiring reliable application programming interfaces (APIs) to execute agentic responses effectively.
Role-based access controls become critical, preventing systems from learning information beyond their authorized scope. Security protocols must ensure that agentic AI systems enhance rather than compromise data protection standards.
Dig deeper:
- Building an AI Team Means Hiring Where the Talent Is, Not Where Your Bank Is
- Four AI Action Plans to Supercharge Your Marketing Team’s Impact
- AI is Changing Customer Interactions. Will Your Bank Join the Conversation?
The Strategic Implementation Playbook
Leading institutions share common approaches that maximize agentic AI’s transformational potential while minimizing implementation risks.
1. Start Simple, Scale Systematically
The most successful deployments begin with targeted applications addressing specific pain points rather than attempting comprehensive transformation. Gupta recommends starting “with a simpler agentic system that addresses a set of the needs within a process and scale that over time to automate bigger and bigger parts of the process itself.”
Promising initial applications include mortgage underwriting automation, small business loan processing, collections management, and “know your customer” compliance. These processes share characteristics that make them ideal for agentic AI: high volume, well-defined parameters, and clear success metrics.
2. Establish Robust Governance Frameworks
HSBC’s approach illustrates comprehensive governance implementation. The institution focuses on five key areas: establishing detailed AI system inventories, prioritizing use cases with clear business value, creating common technology platforms, ensuring data quality and accessibility, and maintaining human-AI collaboration rather than replacement.
The bank’s governance framework emphasizes safety and risk tolerance alignment. “This is all about safety and making sure that our AI systems are being built in a way that we as a firm are comfortable with and within our risk tolerance,” explains Glasner.
3. Focus on Measurable Business Value
Successful implementations tie directly to quantifiable business outcomes. Buluswar at Citi recommends evaluating processes across five dimensions: manual nature, expense level, time delays, error frequency, and consequence severity of mistakes.
This framework helps prioritize high-impact applications. Mortgage underwriting, for example, scores highly across all dimensions—it’s manual, expensive, time-consuming, error-prone, and consequential. Automating such processes delivers clear value that justifies investment and organizational change.
4. Build Platform Capabilities, Not Point Solutions
Rather than developing isolated AI applications, leading institutions create shared platforms supporting multiple use cases. HSBC’s common technology platform allows data scientists, machine learning experts, and software engineers to build systems that are easier to govern, scale, and monitor safely.
This platform approach prevents the chaos that emerges when each team pursues independent AI initiatives. It also enables rapid deployment of new applications as organizational AI capabilities mature.
The Competitive Advantage of Early Adoption
Financial institutions successfully implementing agentic AI are discovering advantages that extend far beyond simple efficiency gains.
Enhanced risk management and security: Agentic AI transforms threat detection by enabling real-time monitoring and response at scales impossible for human teams. These systems automatically identify anomalous behavior patterns, quarantine compromised systems, and manage identity verification processes continuously.
The technology also enables micro-personalization of fraud detection. Banks can customize detection algorithms to individual customer patterns and preferences while maintaining appropriate transparency and user control.
Transformed customer experience: The customer experience implications extend beyond improved response times to fundamentally different service capabilities. Wealth advisors, for example, can leverage agentic AI to synthesize comprehensive client information, tie it to macroeconomic conditions and portfolio observations, then deliver highly targeted financial advice.
This capability enables what Buluswar describes as “much more surgical set of interactions with your clients”—moving beyond generic advice to highly personalized financial guidance based on comprehensive data analysis.
Operational excellence and cost reduction: Back-office automation delivers substantial cost savings while improving accuracy. The technology reduces manual review workloads, freeing human experts for high-complexity, high-value work. Processing times for complex operations like mortgage underwriting can shrink from weeks to days through intelligent automation.
Error reduction and faster error detection create compounding benefits. AI systems can learn from mistakes and improve over time with proper lifecycle management and oversight, creating performance improvements that traditional automation cannot achieve.
The Long-Term Transformation Vision
Looking ahead, banking executives anticipate continued evolution in agentic AI capabilities, with fraud detection (75% priority), security (64%), and customer experience (51%) remaining top development areas.
The technology’s impact will extend throughout banking operations. As Nimish Panchmatia, chief data and transformation officer at DBS Bank, predicts: “I think fundamentally AI will apply to every part of the business: front office, middle office, back office. Fundamentally, the way work gets done is going to be very different in the coming years.”
However, this transformation requires sustained commitment. “Agentic AI is a continuous journey,” emphasizes Panchmatia. “If done properly, there’s significant value at the end of it. But you have to persevere.”
The institutions investing in systematic agentic AI implementation today are positioning themselves for competitive advantages that will compound over time. As these systems learn, adapt, and improve, they create operational capabilities that will be difficult for competitors to replicate quickly.
The banking sector’s rapid agentic AI adoption demonstrates that the technology has moved beyond experimental phase to operational reality. Financial institutions that approach implementation strategically—starting simple, scaling systematically, and maintaining human oversight—are discovering transformational capabilities that reshape both customer service and operational efficiency. The question is no longer whether banks will adopt agentic AI, but how quickly they can implement it effectively to maintain competitive positioning in an increasingly automated financial services landscape.
