Generative AI and Finance

This two-day course gives finance professionals a practical understanding of how Generative AI can be used to research, analyse, communicate and automate everyday financial work. Participants move from understanding LLMs and effective prompting to document-grounded research, data analysis, AI agents, verification and workflow automation, with each theory session followed by an associated hands-on practical session using tools such as ChatGPT, Microsoft Copilot, NotebookLM, Gamma and Codex.

• Portfolio managers, traders and investment professionals
• Investment, research and financial analysts
• Risk, finance and model-validation professionals
• Financial executives, team leads and decision makers
• Anyone looking to use Generative AI productively and responsibly in a finance workflow

• Generate clear hypotheses, scenarios, and investment narratives in minutes.
• Produce high-quality IC memos with risks, triggers, and exit criteria.
• Use GenAI to summarise, synthesise, and challenge investment views
• Improve factuality and reduce hallucinations with simple QA techniques
• Streamline daily work using Copilot in Outlook, Excel, PowerPoint, and Teams
• Build a repeatable decision-making rhythm using GenAI as a structured assistant.

• A working understanding of finance, markets or business analysis
• Basic familiarity with spreadsheets and common office productivity tools
• No prior Generative AI, machine-learning or coding expertise is required
• Participants should have access to an approved GenAI tool; access to Microsoft Copilot, Excel and PowerPoint is useful for the practical exercises

Session 1: Generative AI Foundations, LLMs and Effective Prompting

• What Generative AI and Large Language Models are
• How LLMs generate responses: tokens, context, probability and reasoning
• Strengths and limitations of GenAI in financial work
• Hallucinations and why they occur
• How GenAI differs from traditional search, analytics and automation
• The evolving ecosystem: ChatGPT, Copilot, Gemini, Claude and enterprise AI
• Effective prompting using Role -> Context -> Task -> Constraints -> Checks
• Reusable finance prompting patterns: summarisation, extraction, comparison, hypothesis generation, scenario analysis and decision support
• Confidentiality, security and responsible use in finance

Session 2: Practical Prompting for Financial Research and Decision Making

• Turn news and market information into an executive briefing
• Generate testable investment hypotheses
• Create bull vs. bear arguments
• Challenge an investment view using Devil's Advocate prompts
• Separate facts, assumptions and interpretation
• Convert research into an IC-style memo covering thesis, evidence, risks, catalysts, triggers and exit criteria
• Rewrite the same analysis for a Portfolio Manager, Risk Committee, Management Committee and Client
• Build reusable prompt templates for recurring work

Session 3: AI-Powered Research, RAG and Working with Documents

• General LLM knowledge vs. grounded research
• What Retrieval-Augmented Generation (RAG) is and why it matters
• How AI works with supplied documents
• Why grounding, source attribution and citations matter
• Research across annual reports, earnings transcripts, market research, investment papers, policy documents and internal reports
• Summarisation vs. synthesis vs. analysis
• Comparing views across multiple documents
• Identifying contradictions, gaps and unanswered questions
• Asking AI to distinguish evidence, interpretation, assumptions and missing information
• Turning research into management or investment communication

Session 4: Practical AI Research and Communication

• Use tools such as NotebookLM, ChatGPT/Copilot and Gamma
• Load a financial source pack and ask cross-document questions
• Identify key themes and compare conflicting viewpoints
• Extract supporting evidence and identify missing information
• Produce an executive summary, investment or management briefing, key risks and unanswered questions
• Verify important claims against source material
• Turn the analysis into an executive presentation
• Improve AI-generated slides rather than accepting the first draft

Session 5: From Copilots to Agents - How Modern AI Workflows Work

• Three stages of AI usage: Chat (ask AI), Copilot (work with AI), Agent (delegate to AI)
• What an AI agent is
• Core components: Goal -> Context -> Tools -> Actions -> Memory -> Checks -> Human approval
• How agents use documents, spreadsheets, browsers, email, enterprise systems and code
• Single-step vs. multi-step tasks
• Human-in-the-loop workflows
• When agents are useful and when they should not be used
• Introduction to coding agents such as Codex
• Finance examples: company research, market monitoring, portfolio reporting, management reporting, reconciliation and recurring analysis
• Risks: incorrect actions, excessive permissions, poor data quality and weak controls

Session 6: Practical AI for Excel, Data Analysis and Dashboards

• Give Codex or another AI assistant an Excel/CSV dataset
• Ask AI to understand the data structure
• Clean and transform data using natural languages
• Calculate financial metrics and compare periods
• Identify drivers and anomalies
• Run simple scenarios and generate charts
• Ask AI to explain and show its calculations
• Challenge and validate the output
• Modify the analysis conversationally
• Turn the analysis into a simple Streamlit dashboard
• Example datasets: portfolio analysis, company performance, risk metrics, revenue/margin analysis, market data and scenario analysis

Session 7: Trusting AI - Verification, Hallucinations, Governance and Controls

• Why plausible does not mean correct
• Common failure modes: hallucinated facts, unsupported conclusions, incorrect calculations, misleading citations, confirmation bias and missing context
• The 4 Cs: Context, Citations, Counters and Calibration
• Evidence-first prompting
• Asking AI to show source, calculation, assumption and uncertainty
• Red-teaming your own analysis and challenging the model's first answer
• Creating QA and review checklists
• Confidentiality and sensitive data
• Permissions and access controls
• Human approval points, auditability and traceability
• Responsible AI in regulated financial environments

Session 8: Practical Agents, Automation and the AI-Enabled Finance Workflow

• Focus: Bring everything together into repeatable workflows participants can use in their own roles.