Surendra Reddy Kamasani is CTO at IBM, and a leader in AI, cloud, and digital transformation in financial services.
In today’s competitive financial landscape, success hinges on speed, accuracy and delivering a seamless customer experience. The traditional merchant onboarding process, long regarded as a necessary but cumbersome step, has become a significant bottleneck for banks and fintech companies aiming to scale swiftly and operate efficiently. This process—characterized by lengthy timelines, manual compliance reviews, high operational costs and inconsistent documentation—has often resulted in frustrated merchants, regulatory risks and lost revenue opportunities.
Recognizing these challenges, forward-thinking organizations are turning to AI and automation to reshape and accelerate their onboarding workflows. The journey toward a fully digital, scalable and compliant onboarding system requires a careful combination of advanced technology, strategic planning and enterprise-wide governance.
Over the past several years, I’ve worked closely with enterprise organizations across financial services, insurance and other regulated industries to modernize and transform critical business operations using AI and automation technologies. One of the most impactful initiatives I’ve been involved in focused on modernizing merchant onboarding processes for a large financial payments organization, where I collaborated closely with engineering and architecture teams to help shape an AI-driven, multi-agent onboarding framework with measurable operational goals and phased business outcomes.
Based on those experiences, I’ll explore how integrating a comprehensive AI-driven onboarding framework can provide a realistic, effective blueprint for modern financial institutions.
Challenges In Traditional Merchant Onboarding
Merchant onboarding typically involves a series of manual steps: collecting documentation, verifying identities, evaluating risks and ensuring compliance with anti-money laundering (AML) and know-your-customer (KYC) regulations. These steps are often disjointed, relying on manual reviews and siloed systems, which can lead to several operational inefficiencies:
• Extended Processing Time: I’ve found that onboarding can take anywhere from 30 to 90 days, delaying revenue generation and merchant engagement.
• High Manual Effort: Repetitive tasks such as data verification, document validation and compliance checks often require significant human resources.
• Operational Costs: Manual processes increase staffing needs, result in higher error rates and escalate compliance risks.
• Inconsistent Documentation And Checks: Variability in documentation quality and manual review standards can lead to regulatory exposure and audit difficulties.
The cumulative impact is a poor merchant experience, increased costs and potential regulatory penalties—pressing organizations to seek solutions that streamline and modernize their onboarding workflows.
Leveraging AI And Data-Driven Frameworks
To address these challenges, many organizations are now deploying AI-based frameworks that automate decision making, improve accuracy and accelerate process throughput. Crucially, these frameworks are built upon large data-driven approaches, utilizing Python programming for data analysis, model training and automation scripting.
The core of this transformation lies in developing a suite of specialized AI agents—each designed for a specific function—such as KYC verification, AML screening, risk scoring and compliance checks. These agents are integrated within an orchestration system that coordinates workflows, ensures regulatory adherence and manages exception handling.
Using Python’s extensive data science libraries—such as Pandas, NumPy, Scikit-learn, TensorFlow and language models—developers can build scalable and flexible agent workflows. These workflows can quickly ingest thousands of data points, analyze documents, extract relevant information and make intelligent decisions.
Building The AI-Driven Onboarding Framework
Implementing an AI-powered onboarding system begins with understanding enterprise data and processes at a granular level. Data collection and preprocessing are critical steps; organizations aggregate historical onboarding records, including documents, customer inputs, transaction histories and prior compliance decisions. These datasets inform the training of machine learning models and rule-based engines that form the backbone of the agents.
Developers create modular AI agents:
• The KYC agent uses optical character recognition (OCR) and natural language processing (NLP) to extract data from identity documents, verifying authenticity against known sources.
• The AML screening agent cross-references customer data with established sanction and watch lists, flagging suspicious activity for review.
• The risk scoring agent combines multiple data points—such as transaction behavior, geographic location and business type—to generate an overall risk score.
• The compliance agent ensures that onboarding documentation and decision outputs meet regional regulatory standards through rule-based checks and explainability modules.
These agents operate simultaneously within a multi-agent ecosystem, communicating through an orchestration layer coded in Python that manages workflows, prioritizes cases and escalates issues to human reviewers when necessary.
From Pilot To Scale
Deploying such a system involves phased steps—starting with a pilot project in a specific region or merchant segment. During this phase, organizations evaluate the AI agents’ accuracy, speed and compliance adherence, fine-tuning models based on real-world data and feedback. For example, initial testing may reveal that the OCR module achieves 95% accuracy in document extraction, while the risk scoring agent correctly identifies high-risk cases in 98% of instances.
Over time, with continuous retraining and data collection, these models improve, leading to faster processing times and higher decision precision. This iterative approach enables ongoing alignment with regulatory standards and stakeholder expectations.
Once validated, the system is scaled across regions, with infrastructure provisions designed for elasticity—leveraging cloud platforms such as AWS, Azure or Google Cloud—to handle fluctuating onboarding volumes efficiently.
Governance, Explainability And Compliance
Implementing AI at scale requires rigorous governance frameworks. Key components include:
• Model Explainability: Incorporating techniques like LIME or SHAP provides transparency into AI decisions, satisfying regulatory audit requirements.
• Bias And Drift Monitoring: Continuous evaluation detects model degradation or bias shifts, prompting retraining to maintain fairness and accuracy.
• Audit Trails: Detailed logging of every decision point builds a comprehensive audit trail, enhancing accountability and regulatory compliance.
• Regulatory Alignment: Maintaining an up-to-date knowledge base of regional laws and integrating rule-based checks enables ongoing compliance.
Real Impact And Future Outlook
The results of deploying this AI-powered onboarding framework are compelling. In a project I’m currently working on using this framework, I’ve been able to reduce onboarding timelines from 30 to 45 days to just a few hours or minutes. Manual review efforts have also decreased by up to 70%, lowering operational costs while improving accuracy and consistency. Additionally, enhanced compliance controls significantly reduce regulatory risks.
The future of financial services lies in leveraging advanced AI frameworks to foster operational excellence, regulatory compliance and superior customer engagement. By integrating data-driven methods, robust Python workflows and intelligent agentic systems, organizations can transform their merchant onboarding processes—moving beyond traditional bottlenecks to a dynamic, scalable and intelligent digital frontier.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
