AI-Driven Process Automation For A Financial Institution

The client is a Netherlands-licensed bank with an international footprint across seven countries. The bank supports more than 400,000 retail and corporate customers worldwide. In the financial sector, the organization manages complex internal processes that include trade documents, compliance checks, KYC workflows, and legacy internal systems. Many of these workflows require accurate document handling, fast access to internal knowledge, strict control over sensitive data, and reliable operational decision-making. The project focused on helping the bank introduce AI into internal operations in a structured, scalable, and governed way. The work combined AI consultancy and strategic enablement with the delivery of practical AI-powered applications for everyday banking workflows.

AI consultancy and strategic enablement

• Provided strategic AI consultancy across internal banking operations, identifying where AI could deliver measurable value.
• Evaluated AI opportunities, adoption risks, and organizational readiness; identified feasible use cases and turned broad ideas into internal solutions.
• Defined and supported a long-term AI strategy aligned with business, product, operational, and compliance priorities.
• Identified and escalated organizational blockers affecting AI delivery, adoption, and scalability.
• Presented AI initiatives, architectural decisions, roadmap milestones, and transformation progress to stakeholders and leadership.
• Participated in cross-company AI knowledge-sharing sessions to exchange best practices around AI architecture, governance, SDLC transformation, tooling adoption, and operational workflows.

AI foundation architecture and governance

• Designed an internal AI foundation as a reusable base, establishing standardized architectural patterns, tooling, and shared AI capabilities applicable across multiple internal projects.
• Architected governance-aware AI frameworks with built-in auditability, observability, and decision traceability, covering operational oversight, risk management, compliance, and lifecycle control.
• Embedded governance rules directly into the AI framework’s code, making compliance enforceable by design rather than by manual review.
• Added automated governance checks to the deployment pipeline to catch rule violations and unsafe configurations before they reached production.
• Adapted governance frameworks to product-specific requirements, including PII handling, human-in- the-loop workflows, and access control.

AI-assisted engineering and SDLC transformation

• Designed SDLC-focused AI agents to improve engineering productivity, development workflows, and software delivery processes.
• Evaluated and supported the adoption of AI-assisted engineering tools, including Claude, GitHub Copilot, and OpenCode.
• Established internal best practices and usage guidelines for AI tooling adoption.
• Designed custom Agile operating frameworks for AI and R&D teams where traditional Scrum was not sufficient.
• Introduced separate Discovery and Delivery tracks to improve transparency for management and business stakeholders.
• Improved handling of R&D capacity planning, spikes, undefined research workflows, and experimentation management.

AI-powered document processing application

• Created an AI-powered document processing application for trade documents, invoices, shipping documents, and other scanned banking materials.
• Used OCR and Azure Document Intelligence to convert PDFs and scanned files into structured data.
• Tested several OCR and document intelligence options before selecting the most reliable approach.
• Designed the solution to extract entities and related metadata, including countries and contextual document information.
• Converted extracted information into JSON for structured processing.
• Created a user interface where employees could review the extracted data and select the required fields.
• Focused on extraction precision because the solution was intended for everyday banking workflows involving sensitive and operationally important documents.

AI-assisted legacy code modernization

• Created an AI-assisted tool to translate a legacy frontend codebase into a newer internal framework.
• Designed the solution around LLM context management, which was one of the key technical challenges in converting a large legacy application.
• Built an MCP server with dedicated tools, enabling the AI agent to work with project-specific context rather than producing isolated code suggestions.

AI-powered employee request answering system

• Created an internal AI answering system for employees handling customer personal data requests.
• Built the solution from scratch, including the user interface, backend, and AI functionality.
• Created a flow where employees could upload emails and receive generated response drafts based on approved company knowledge sources.
• Designed the system to search the internal knowledge base and retrieve relevant context before generating an answer.
• Addressed the challenge of large knowledge base management, including document lookup, context retrieval, and LLM response generation.

AI-powered adverse news monitoring solution

• Built an AI-powered solution to scan the web for adverse news related to financial business partners.
• Designed the system to search across public sources, including news websites, articles, government sources, and trade websites.
• Added domain filtering and source management to improve result quality.
• Created a media/source database that could be updated as new relevant sources were discovered or irrelevant sources needed to be removed.
• Implemented the solution in two formats: a self-managed application and a PaaS-based version using Azure Logic Apps.
• Delivered the first version from scratch in approximately 1 week.

• Reduced document processing time by approximately 50% through AI-powered OCR and structured data extraction.
• Accelerated legacy frontend modernization from an estimated 1 to 2 years to approximately 4 to 5 months, delivering up to 80% development time savings.
• Delivered an AI-powered employee request answering MVP in approximately 4 days.
• Provided an adverse news monitoring solution in approximately 1 week.
• Established a structured AI adoption model, helping the client access missing AI expertise and move from fragmented ideas to scalable, governed implementation.
• Created a reusable AI foundation to support future AI-driven products, internal automation initiatives, and enterprise-grade governance.
• Improved operational efficiency by automating document-heavy, compliance-related, and knowledge- based internal banking workflows.