As the financial services sector continues its aggressive digital transformation, agentic artificial intelligence has rapidly ascended to the forefront of industry discourse. Promoted as the next evolutionary leap beyond traditional generative AI and large language models, agentic systems possess the capability to autonomously plan, execute multi-step workflows, and make decisions to achieve complex business goals. However, as financial institutions navigate the fine line between technological innovation and rigorous risk management, a critical question remains: how much of the current enthusiasm is translating into measurable enterprise impact?
To address this pressing industry inquiry, GlobalData recently convened a panel of prominent technology and financial sector experts for an exclusive roundtable discussion. Entitled “Agentic AI in Banking & Financial Services: Beyond the Hype,” the virtual event brought together leading minds to dissect the realities of deploying autonomous AI agents in heavily regulated financial environments. Among the featured speakers were Jeff Veis, Chief Marketing Officer at Impetus Technologies; Deepak Khosla, Chief Growth Officer and Head of AI Business at Impetus Technologies; and Stephen Walker, Retail Banking Analyst at GlobalData.
The comprehensive dialogue provided attendees with a grounded assessment of the current state of agentic AI adoption, moving past the marketing gloss to examine the profound technical, operational, and regulatory hurdles facing modern banks, asset managers, and insurance providers.
The Current Landscape: Experimentation Versus Scale
The financial services industry has historically been an early and aggressive adopter of advanced computing technologies, ranging from high-frequency algorithmic trading engines to cloud-based data lakes. Consequently, the advent of generative AI in late 2022 sparked immediate interest across the banking sector. Financial institutions rushed to implement conversational chatbots, automated document summarization tools, and coding assistants to boost back-office productivity.
However, the transition from passive generative AI tools—which simply respond to user prompts with text or code—to active agentic AI systems has proved considerably more complex. While a significant majority of tier-one and tier-two financial institutions are currently running proofs of concept (PoCs) or small-scale pilot programs utilizing agentic frameworks, a relatively small fraction have successfully deployed these systems into mission-critical production environments.
During the GlobalData roundtable, Stephen Walker, Retail Banking Analyst at GlobalData, highlighted the widening gap between experimentation and enterprise-wide deployment. According to industry observations discussed during the session, financial institutions are eager to harness the efficiency gains promised by autonomous agents, yet they find themselves constrained by legacy technological debt, siloed data repositories, and stringent regulatory oversight.
The Paradigm Shift: Moving Beyond Document Summarization
To understand why agentic AI deployment differs so sharply from previous technological implementations, industry leaders emphasize the unique nature of autonomous agents within a financial context. Deepak Khosla, Chief Growth Officer and Head of AI Business at Impetus Technologies, articulated this distinction during the roundtable discussion.
“An agent in banking is not just summarizing a document,” Khosla stated. “It could influence and impact credit, fraud, payments, customer treatment, reporting, or advice. The bar for production is therefore much higher in the banking and financial services sector.”
Unlike standard retrieval-augmented generation (RAG) applications that simply fetch and display information for human review, true agentic workflows empower software to take independent actions. An autonomous financial agent might be tasked with investigating suspicious transaction patterns, recalculating risk profiles in real-time during market volatility, or autonomously initiating workflows to onboard corporate clients while adhering to strict Anti-Money Laundering (AML) and Know Your Customer (KYC) regulations.
Because these actions carry direct financial, legal, and reputational consequences, the margin for error is virtually nonexistent. A hallucinating chatbot in a retail setting might provide incorrect store hours; a malfunctioning agent in a commercial banking environment could erroneously approve high-risk credit lines or trigger erroneous multi-million-dollar wire transfers.
Key Barriers to Production Adoption
The roundtable participants conducted a deep dive into the primary impediments slowing the deployment of agentic AI at scale. They categorized these obstacles into several interconnected domains: data readiness, fragmented enterprise context, governance gaps, safety risks, and evolving regulatory expectations.
Data Readiness and Infrastructure Deficits
For an AI agent to function effectively, it requires access to pristine, structured, and unstructured data across the entire enterprise ecosystem. Unfortunately, decades of mergers, acquisitions, and legacy system patching have left many traditional financial institutions with deeply fragmented data architectures. Customer information, transaction histories, credit scoring models, and compliance logs often reside in isolated silos that do not communicate seamlessly with one another.
Fragmented Enterprise Context
Building upon the data readiness challenge, Khosla emphasized that an agent is only as intelligent as the context it can access. In many institutions, this context is stale, poorly governed, or incomplete. Without a unified view of a customer’s relationship with the bank—spanning retail accounts, mortgage products, wealth management portfolios, and commercial loans—an autonomous agent cannot make holistic, informed decisions.
Governance and Safety Risks
Financial institutions operate under some of the strictest governance frameworks in the global economy. Deploying autonomous software introduces unprecedented security vulnerabilities, including prompt injection attacks, unauthorized data access, and unpredictable agent loops where autonomous systems make recursive decisions that drift far outside intended operational parameters. Ensuring that human-in-the-loop safeguards are properly calibrated without neutralizing the efficiency gains of automation remains a central design challenge.
Regulatory Compliance and Supervisory Expectations
Regulators worldwide—including the US Federal Reserve, the Office of the Comptroller of the Currency (OCC), the European Banking Authority (EBA), and the UK Financial Conduct Authority (FCA)—have made it clear that financial institutions remain fully accountable for the decisions made by automated systems. The “black box” nature of complex neural networks conflicts directly with regulatory mandates requiring transparency, explainability, and fairness in lending, pricing, and consumer treatment.
The Solution: Engineering Enterprise Context
Recognizing these formidable barriers, the experts during the GlobalData roundtable pivoted toward actionable strategies for overcoming them. Rather than waiting for silver-bullet artificial general intelligence (AGI) models to solve these problems natively, the panel argued that financial institutions must take an active, engineering-led approach to data and context preparation.
Successful agentic AI adoption, according to Impetus Technologies and GlobalData, relies heavily on building robust, AI-ready data foundations. This foundational work involves grounding AI agents and underlying systems in proprietary business processes, operational realities, historical client interactions, and institutional governance policies.
“Our approach starts with the belief that agentic AI success depends on the quality of enterprise context available to agents,” Khosla explained during the event. “If that context is fragmented, stale or poorly governed, they will fail in production.”
To achieve this high standard of context quality, the session concluded that financial institutions should prioritize context engineering. This discipline involves constructing trusted semantic layers, comprehensive enterprise ontologies, and sophisticated knowledge graphs that AI agents can reliably traverse and reason over. By translating raw, chaotic database tables into structured relational maps infused with business logic, banks can provide their autonomous agents with the precise navigational tools required to operate safely and effectively.
Targeting High-Impact, Measurable Use Cases
As financial technology budgets face heightened scrutiny in an uncertain macroeconomic climate, institutions are increasingly demanding clear, quantifiable returns on investment (ROI) for any artificial intelligence initiative. The roundtable participants advised financial leaders to avoid chasing overly broad, generalized AI use cases and instead focus their agentic AI strategies on targeted, high-impact domains where risk can be rigorously managed and ROI is easily measured.
Prominent high-impact use cases discussed included:
- Enhanced Fraud Detection and Prevention: Deploying autonomous agents capable of continuously monitoring transaction streams, correlating disparate fraud signals across channels, and initiating defensive holds with minimal false positives.
- Intelligent Regulatory Reporting: Utilizing agents to automatically aggregate data from diverse business units, reconcile discrepancies, and draft compliance filings under tight regulatory deadlines.
- Streamlined Commercial Loan Origination: Empowering agents to gather initial documentation, perform preliminary credit risk assessments, and draft underwriting memos, thereby reducing loan processing times from weeks to days.
- Wealth Management and Client Advisory Support: Providing human advisors with autonomous research assistants that synthesize market intelligence, portfolio performance metrics, and client life-event triggers to generate hyper-personalized advisory insights.
Broader Industry Implications and Future Outlook
The insights emerging from the GlobalData roundtable reflect a broader maturation phase within the financial technology sector. The initial frenzy of excitement surrounding generative and agentic AI is steadily giving way to pragmatic, engineering-focused implementation strategies.
Financial institutions are realizing that deploying agentic AI is not merely a software upgrade, but an organizational transformation that demands alignment across Chief Information Officers (CIOs), Chief Risk Officers (CROs), Chief Data Officers (CDOs), and business line leaders. Those institutions that successfully build out their semantic data layers, establish rigorous guardrails, and master context engineering will likely capture significant competitive advantages in operational efficiency, customer experience, and risk management.
Conversely, institutions that rush unvetted autonomous agents into production without adequate data foundations risk severe operational failures, regulatory penalties, and reputational damage. As the industry moves further into 2026 and beyond, the differentiator between AI leaders and laggards will not be the sophistication of the foundational models they license, but the quality of the proprietary enterprise context they feed into those models.
For financial services executives, risk managers, and technology leaders seeking a deeper understanding of these dynamics, the full GlobalData roundtable session, “Agentic AI in Banking & Financial Services: Beyond the Hype,” remains available for viewing online, offering further technical guidance on how context engineering can enable the safe and effective deployment of autonomous systems across the enterprise.
