Online-Development Programme: AI for Banking and Finance: Analytics, Local Models, and Custom AI Solutions

• Status:The new season starts on November 23, 2026, 5 sessions of 2 hours each
• Duration:23 November - 07 December
• Language: Ukrainian
• Time:17:00–19:00, every Monday and Wednesday: November 23, 25, 30 and December 02, 07 Kyiv time
• Fee for EBA members:11500 UAH
• Early Registration for EBA members:10000 UAH until October 23, 2026
• Place: Will be online
Cancel registration Register Register a colleague

About the Programme

AI in banking and finance has long gone beyond ChatGPT and text generation.

Language models can analyze spreadsheets and documents, identify deviations in financial indicators, work with internal knowledge bases, help create reports, and even write code for small internal applications.

At the same time, there is an important aspect to consider.

In the financial sector, you cannot simply upload an internal document to the first available chat and hope that everything will be fine. You need to understand where the data is processed, which model is being used, what is stored in the cloud, and what can be run locally.

That is why the program will combine working with LLMs, financial analytics, local models, knowledge bases, and vibe coding.

In other words, we will not simply study individual tools. We will take a specific task and build a solution around it.

EBA Management Development Centre Team, together with Olena Chetyrina, invite you to participate in the new Online-Development Programme: AI for Banking and Finance: Analytics, Local Models, and Custom AI Solutions.

Target audience:

• Financial analysts.
• Employees of banks and financial institutions.
• Heads of finance departments.
• Risk, compliance, and internal control specialists.
• Specialists who work with reporting, documents, and large volumes of data.
• Teams planning to implement AI solutions in internal processes.

After the program, participants will be able to:

• Select the right model for a specific financial task.
• Distribute tasks between cloud-based and local models.
• Run open-source models locally.
• Analyze spreadsheets, reports, and documents using AI.
• Design AI assistants based on internal documents.
• Understand how data should be stored for AI systems.
• Create prototypes of internal financial applications using AI.
• Validate model outputs and determine where specialist oversight is required.

Programme:

Module 1. LLMs for Banking and Financial Tasks – November 23

LLMs are no longer just chat tools that can be asked to write an email or shorten a document.

A model can be connected to a database, internal documents, a website, an analytics system, or a custom application. But first, it is important to understand the difference between the model itself, an AI platform, an API, and a ready-made service.

We will discuss:

• How language models work and why they can make mistakes.
• The differences between ChatGPT, Claude, Gemini, open-source models, and local solutions.
• Which models work best with text, analytics, documents, logic, and code.
• What a context window is and why a model may lose part of the information.
• How to reduce hallucinations.
• When a cloud-based model can be used and when a local environment is required.
• How to avoid overpaying for a complex model when a simpler one can handle the task.

Practice:

Participants will map their own tasks and determine:

• What data is used.
• How confidential it is.
• Which model can be connected.
• Where local processing is required.
• How the result should be validated.

Module outcome:

A ready-to-use matrix for selecting AI solutions for the tasks of their department.

Module 2. Local Models and Working with Confidential Documents – November 25

A local model runs on a computer or an internal server. Data does not need to be transferred to an external AI service.

However, simply installing Ollama does not automatically make the system secure. Access controls, file storage, internet connectivity, logs, and the application architecture must also be properly managed.

We will discuss:

• What local and open-source models are.
• How to choose a model based on the task and computer specifications.
• Which models can be used for text, documents, analytics, and reasoning.
• How to work without an internet connection.
• How to evaluate the quality of a local model’s responses.
• When a local model is sufficient and when more advanced enterprise infrastructure is required.

Practice:

We will run a local model and test it on an anonymized financial document:

• Formulate a prompt.
• Extract structured data.
• Verify the facts.
• Compare the result with a cloud-based model.

Module outcome:

A configured local model and an understanding of which work tasks it can be used for.

Module 3. Financial Analytics with AI – November 30

A good-looking dashboard does not necessarily mean that a sound decision can be made based on it.

First, you need to understand where the data is stored, what needs to be calculated, which indicators should be compared, and what each deviation means. Only then should you start building charts.

We will discuss:

• How to conduct an inventory of data sources.
• How to work with Excel, CSV, PDF, CRM, ERP, and other internal systems.
• How to clean, structure, and combine data.
• How to analyze revenue, expenses, trends, and deviations from plan.
• How to identify anomalies and unusual transactions.
• How to conduct scenario analysis.
• How to verify formulas and calculation logic.
• How to create custom dashboards and analytical interfaces using AI.

Practice:

Using an anonymized dataset, participants will:

• Identify the required indicators.
• Prepare the data.
• Conduct the analysis.
• Identify deviations.
• Formulate a management conclusion.
• Create a simple visualization.

Module outcome:

A ready-to-use financial analysis flow that can be adapted to a specific task and data sources.

Module 4. Local AI Assistant with an Internal Knowledge Base – December 02

Uploading 200 documents into a chat and calling it a knowledge base is not enough.

Documents need to be cleaned, structured, split into chunks, and supplemented with sources, dates, versions, and access permissions. Only after that should the model be connected.

We will discuss:

• How an assistant with its own knowledge base works.
• What RAG, embeddings, chunks, and semantic search are.
• Which data should be stored in tables and which in documents.
• How to organize search across policies, instructions, procedures, and methodologies.
• How to make the assistant respond based on the provided sources.
• How to add references to the document, page, date, and version.
• How to organize information updates and access control.
• How to test the knowledge base and identify errors in responses.

Use cases:

• An assistant based on internal procedures.
• A compliance assistant.
• Search across regulatory documents.
• An assistant for banking products.
• An onboarding assistant for employees.
• A system for checking document completeness.

Practice:

We will design a local assistant that:

• Works with internal documents.
• Finds the relevant fragment.
• Generates a response.
• Shows the source.
• Indicates when there is not enough information in the knowledge base.

Module outcome:

The architecture of a local assistant and prepared requirements for the knowledge base.

Module 5. Creating a Custom Financial Application with Vibe Coding – December 07

Vibe coding makes it possible to create digital solutions through text prompts: applications, dashboards, calculators, knowledge bases, automations, and internal services.

However, the same rule applies here: if the task is described vaguely, AI may generate a large amount of code without a clear understanding of what it actually does.

That is why we will start with the logic of the solution.

We will discuss:

• How to turn a business task into a technical specification.
• How to describe users, data, functions, and constraints.
• How to create interfaces and code with AI.
• How to connect a local model.
• How to work with files and databases.
• What CRUD, authorization, and access control are.
• How to store keys securely and avoid exposing them in open code.
• How to verify calculations and test an application.
• What needs to be considered before moving a prototype into real-world use.

Prototypes:

• A financial document analyzer.
• A management report generator.
• A spreadsheet reconciliation tool.
• A credit file completeness checker.
• A local financial calculator.
• A dashboard for performance analysis.
• A search tool for internal documents.

Practice:

Participants will build a prototype of a local application using synthetic or anonymized data.

In other words, we will go through the entire process: from defining the task to building the interface, logic, model integration, and initial testing.

Module outcome:

A working prototype and a plan for its further development and implementation.

Please note that any opinions and views expressed by the trainer(s) during the event are solely their personal views and intellectual property and do not necessarily reflect the official position of the European Business Association.

You can send a question to: