AI Agents' Real Deployment in Banks: From Compliance Screening to Process Reengineering

The Real Implementation of AI Agents in Banking: From Compliance Screening to Process Reengineering

Financial crime compliance is one of the most heavily regulated areas in banking. As transaction volumes grow tenfold, the traditional manual screening model is reaching its limits. Fintech companies such as WorkFusion and Genpact are pushing AI agents into production environments—the former has achieved over 10 implementation cases among the top 20 banks, while the latter emphasizes that AI agents must understand the context of business processes to create real value.

Industry Background

The compliance pressure on the banking industry continues to escalate. The number of alerts generated in areas such as anti-money laundering, sanctions screening, and know-your-customer (KYC) far exceeds manual processing capacity. According to WorkFusion CEO Adam Famularo, if a large bank relied solely on human resources to meet regulatory requirements, it would need to hire approximately 1,800 additional compliance analysts, with extremely high training and retention costs. At the same time, the trend toward real-time payment systems forces banks to complete screening within shorter timeframes, making traditional batch processing methods inadequate.

Current Development Dynamics

WorkFusion: AI Agents Move from Pilots to Large-Scale Production

WorkFusion (acquired by UiPath) focuses on AI agents for financial crime compliance. Its product "Evan" is specifically designed for adverse media screening and once completed scanning 80 million entities in a single day. Famularo stated that AI agents are no longer stuck in the pilot phase: "We have more than 10 of the world's top 20 banks that have formally deployed them in the compliance domain."

Core application scenarios include sanctions list screening, customer name matching, and adverse media screening. In these highly repetitive, high-volume tasks, AI agents can automatically handle 50-70% of false positives, allowing analysts to focus on more complex cases. In fraud investigation and KYC, AI agents play a supporting role, helping analysts speed up their work by simplifying information retrieval.

Famularo also mentioned a cultural innovation: WorkFusion gives its AI agents names, faces, and identities, helping banks view them as "employees" rather than tools. This design makes it easier for banks to accept AI agents as part of the team rather than a simple replacement.

Genpact: AI Agents Need to Understand Workflow Context

Unlike WorkFusion's focus on specific tasks, Genpact approaches from the perspective of enterprise business process services. CEO Balkrishan Kalra pointed out that the current AI agent market has "a lot of noise," but the real differentiator lies in the ability to handle the "last mile" of enterprise operations—the exceptions, handoffs, and judgment calls that cannot be covered by standardized software or publicly available training data.A report jointly released by Genpact and HFS Research shows that 85% of corporate executives believe technology debt, data debt, process debt, and talent debt limit the realization of AI value, and only 6% of enterprises can effectively repair these debts. Vijay Vijayasankar (Genpact's Global Chief AI Officer) emphasizes that models themselves rarely work directly in enterprise environments; a suitable governance framework must be established around the model.

Genpact's strategy is to start from its core operational areas—such as finance, procurement, supply chain, and insurance—embedding AI agents into specific processes like invoice processing, claim review, and procurement approval. Elena Christopher (Genpact's Vice President of Strategic Programs) warns that if the process itself is flawed, AI agents may "execute wrong steps faster without supervision."

Impact on the Financial System

Compliance Efficiency and Cost

The application of AI agents in financial crime compliance directly reduces the labor cost of handling false positives. Famularo points out that 70% of false positives can be fully adjudicated by AI agents, meaning banks can focus senior analysts' attention on truly high-risk cases. This efficiency improvement indirectly affects the speed of payment settlement, as compliance screening is often a bottleneck in the payment process.

Financial Inclusion

Compliance costs are one of the main barriers for banks to serve the lower-end market. The automation capabilities of AI agents are expected to reduce the per-customer compliance cost, making banks more willing to serve low-margin scenarios such as small accounts and cross-border remittances. Over the long term, this could promote financial inclusion.

Banking Competition Landscape

Banks that adopt AI agents first will gain competitive advantages in compliance costs, response speed, and risk control. Small banks may acquire similar capabilities by purchasing SaaS services, thus narrowing the technology gap with large banks. However, the Genpact report also points out that banks with severe process debt may struggle to gain returns from AI investments.

Challenges

Model Trustworthiness and Transparency

Famularo says that banks need to see AI agents running in their own systems before building trust. WorkFusion's approach allows banks to oversee the decision logic of the model and train it to adapt to specific bank strategies. However, the entire industry still lacks a unified standard for AI explainability.

Data Privacy and Regulatory Uncertainty

AI agents need access to large amounts of customer transaction data, raising data privacy concerns. Regulators in different countries have varied attitudes toward the use of AI in financial compliance, with some requiring manual review of all AI decisions, increasing implementation complexity.

Technology Integration and Process Debt

Genpact's research indicates that most enterprises have significant technology and process debt. Layering AI agents on top of legacy systems may complicate the problem. Banks need to undertake process redesign and data governance before deploying AI.## Future Outlook

Famularo believes that as AI and real-time payment technologies increase transaction volumes, the banking industry will face more fraud and financial crime activities, and the use of AI agents will continue to expand. "We will only find more ways to use technology to stop bad actors." He predicts that AI agents will extend from the compliance domain to broader banking functions such as credit assessment, customer service, and risk management within the next three to five years.

Genpact emphasizes that the success of AI agents depends on a deep understanding of the business. The future winners will be those who can both harness AI technology and redesign processes. The industry needs to shift from "automating existing processes" to "designing processes for AI."

Overall, AI agents are moving from proof of concept to large-scale production, but banks' technological readiness and regulatory frameworks still need to evolve in parallel. Over the next five years, AI agents in the compliance field may become the standard, while more general agentic AI will gradually permeate bank operations.

*This article is compiled based on Newsweek AI Impact weekly report, with editorial adjustments.*