Artificial Intelligence in Finance - Fundamentals

Finance is digital and the Financial Services Industry acknowledges the need for a thorough digital transformation as the only means to thrive in the future. Technological capabilities are essential for a future in an industry that is digital in its very essence, the times of managing physical money and bonds being long gone.

Mastering the development and realisation of innovative financial products and services through digital technology is key. Financial systems thus become more reliable and transparent, and user interactions smoother. In terms of user-friendliness and adding value to said interactions, cybersecurity, authentication, (mobile) payments, robo-advisors, etc. all require adequate integration and packaging. 

This course will guide you through the core concepts, techniques, and algorithms that are reshaping financial services. Participants will explore how AI can optimise operations, enhance decision-making, and transform customer interactions. With a focus on practical applications, the training will cover predictive analytics, machine learning models, and neural networks, equipping you with the knowledge to implement AI solutions effectively within your financial organisation.

After completion of the 24-hour course, participants will be able to:

• Explain core AI, machine learning, deep learning, generative AI, and agentic AI concepts in financial services.

• Identify relevant banking use cases in credit scoring, fraud detection, AML support, risk management, compliance, trading support, operations, and customer interaction.

• Understand how Model Context Protocols (MCPs) can connect AI applications to banking tools, data sources, documents, workflows, and controlled actions.

• Design a basic safe AI banking agent using tool allowlists, least privilege, role-based access, human approval, logging, and escalation paths.

• Recognise core risks: hallucination, overreliance, bias, prompt injection, indirect prompt injection, data leakage, excessive agency, insecure tool/plugin design, and third-party MCP risk.

• Apply secure-by-design controls: data minimisation, masking, authentication, authorisation, secrets management, sandboxing, auditability, monitoring, and incident response.

• Prepare a simple AI governance artefact: use-case description, risk assessment, model/tool inventory, control checklist, testing plan, and deployment recommendation.
  

The current foundations are retained and expanded with MCPs, secure AI-agent architecture, threat modelling, testing, monitoring, and a banking capstone.

Module 1: AI in finance - business context and transformation - AI, ML, deep learning, generative AI, agentic AI; banking value chain; practical limits in regulated environments. Activity: map AI opportunities across front office, risk, compliance, operations, IT, and customer service.

Module 2: Data, models, and machine learning fundamentals  - Supervised, unsupervised, and reinforcement learning; financial datasets; feature engineering; training, validation, testing; false positives, false negatives, bias, and drift. Mini-lab: evaluate a simple classification model.

Module 3: Predictive analytics, credit scoring, and risk management  - Creditworthiness, probability of default, early-warning indicators, explainability, adverse-impact concerns, human override, and audit trails. Exercise: assess an AI-supported credit decision workflow.

Module 4: Fraud detection, AML support, and anomaly detection - Rule-based versus ML-based fraud detection; transaction monitoring; anomaly detection; alert triage; investigator support; controls against autonomous closure of alerts. Exercise: design a fraud-alert assistant.

Module 5: Generative AI in banking - LLMs, embeddings, retrieval-augmented generation, document Q&A, summarisation, classification, chatbots, policy assistants, hallucination controls, source grounding, and confidential-data handling.

Module 6: AI ethics, governance, and regulation - Fairness, accountability, transparency, explainability, privacy, GDPR, EU AI Act awareness, DORA operational resilience, outsourcing/third-party risk, and management accountability. Exercise: one-page AI use-case assessment.

Module 7: Introduction to AI agents for banking - Agent components: model, system instructions, memory, tools, planner, policy layer, logs, and human approval. Banking examples: relationship-manager assistant, KYC reviewer, compliance copilot, fraud investigator assistant, and internal service-desk agent.

Module 8: MCP fundamentals for financial institutions - MCP concepts: host, client, server, tools, resources, prompts; connecting AI systems to documents, databases, case systems, product catalogues, and controlled workflows. Activity: define safe MCP tools and permissions.

Module 9: Secure-by-design banking agent architecture - Least privilege, RBAC/ABAC, deny-by-default tool access, read-only/draft-only/approval-required action tiers, identity boundaries, data minimisation, masking, secrets management, and separation of customer-facing versus employee-facing agents.

Module 10: Threat modelling AI agents and MCP integrations - Prompt injection, indirect prompt injection, tool abuse, excessive agency, sensitive-information disclosure, malicious or compromised MCP servers, supply-chain risk, data exfiltration, and identity confusion. Activity: threat-model a banking operations assistant.

Module 11: Testing, monitoring, and red teaming banking agent - Factuality tests, policy-compliance tests, prompt-injection tests, data-leakage tests, tool-call safety tests, regression testing, red-team prompts, monitoring, alerts, incident response, and prompt/model change management.

Module 12: Capstone - safe MCP-enabled banking agent - Participants design an internal relationship-manager assistant that can retrieve approved product information, summarise permitted customer context, check policies, draft notes and follow-ups, escalate issues, and require approval for sensitive actions. Deliverables: architecture, MCP tool list, data-access matrix, control register, test plan, monitoring plan, and go/no-go recommendation.

Suggested timetable:

Day 1: AI foundations, data, ML, credit scoring, fraud, AML support, and financial-risk use cases.

Day 2: Generative AI, governance, ethics, regulation, banking controls, and practical use-case assessment

Day 3: MCPs, safe AI agents, secure architecture, threat modelling, testing, monitoring, and capstone.  

• Banking and financial-sector professionals who need to understand AI opportunities, limitations, risks, and controls.

• Risk, compliance, AML, audit, operations, digital transformation, product, IT, data, and innovation teams.

• Managers and subject-matter experts who may sponsor, review, govern, or test AI and GenAI use cases.

• Technical and semi-technical participants involved in designing or supervising AI assistants, MCP integrations, or agentic workflows.

No coding background is mandatory, but basic digital literacy and familiarity with banking processes are recommended.