AI & Data Science
The AI & Data Science module covers all CNB innovations in the area of artificial intelligence and its testing and application in day-to-day operations, with a focus on streamlining and simplifying routine tasks, freeing up expert capacity for strategic, conceptual and complex assignments and ensuring security and the protection of sensitive data. It also includes innovative projects in the area of economic data analysis.
Introducing AI at the CNB
Artificial intelligence has enormous potential to make routine and time-consuming work easier for people. It enables them to devote more time to creative activities and strategic thinking. As a result, it can open up entirely new possibilities across many fields of human activity – including (central) banking.
AI also has its place at the CNB. In 2024, the CNB purchased its first several dozen licences for the generative AI chatbot, ChatGPT Team. Since then, selected employees have been testing AI capabilities and practical applications (see, for example, the blog article First use of AI in inflation forecasting at the CNB) and proposing ideas on how AI can help fulfil everyday tasks more efficiently, analyse data and support innovation.
“A great idea for using artificial intelligence was born during one of my breakfast meetings with CNB staff. Since then, we’ve launched pilot AI projects at the bank,” says CNB Governor Aleš Michl, who believes that programming with the help of AI will be key.
A Coordination Group for AI also operates within the CNB. This is a project team made up of representatives from departments across the CNB. Its main task is to define AI-related needs, coordinate implementation and share best practices across departments. The group ensures the effective rollout of AI within the central bank, eliminates duplicate initiatives and connects individual departments with the IT team for technical support. The coordination team also determines the framework and direction for AI initiatives across the bank to ensure that the activities of other teams are coordinated and contribute to a unified strategy.
- AI Coordination Team – made up of representatives from all departments. Its task is to define AI-related needs, coordinate implementation and share best practices.
- IT team for AI – provides technological support, cyber security and integration of AI into systems.
- HR team for AI – focuses on training, skills development and employee support in adopting AI.
The results of a CNB employee survey conducted in April 2025 show that 81% of respondents use artificial intelligence at least once a week, with more than one-third (34%) reporting daily use. Only 18% of respondents indicated that they do not yet use AI tools.
The survey also confirmed a strong willingness among employees to contribute to further AI development within the CNB. More than half of respondents (56%) expressed a direct interest in participating in testing, while another 36% said they would consider getting involved if they had more information.
In mid-2025, the CNB took another significant step towards broader use of artificial intelligence by introducing the new Copilot Chat tool for all employees. Enterprise licences for ChatGPT were also acquired to test advanced generative AI capabilities and support tasks across organisational units.
In addition, the CNB began testing other AI tools – for example, Claude by Anthropic, which expands capabilities in data analysis and process automation, and specialised tools in the field of legislation. The Llama language model is also available in the data analysis environment known as DSLab (for more details see below). The aim is to verify the practical benefits of different approaches to AI, ensure their safe use and gradually establish a comprehensive framework for the responsible and effective application of AI within the central bank.
Examples of specific AI applications in practice include not only the First use of AI in inflation forecasting at the CNB, but also the use of AI to monitor financial influencers and accelerate the development of economic models (for example, when transcribing mathematical equations, generating documentation and creating visualisations in the form of diagrams).
Data Science Laboratory
The Data Science Laboratory (also referred to as DSLab or the CNB Shared Data Laboratory) is a shared environment for conducting data analysis and economic research, as well as performing advanced analytical machine learning/AI tasks using Python, R and Matlab.
- The environment was made available to CNB employees in its initial phase in December 2024.
- Logical separation of computing capacity for individual users is ensured through container technology (Podman).
- Shared group drives and the GitLab versioning system allow users working on the same tasks to share code and data.
- Thanks to centralised administration and regular software security scans, DSLab provides a guaranteed, supported and secure solution for working with data analysis languages (Python/R/Matlab).
In 2025, the Data Science Laboratory environment was expanded with several innovative features, for example:
Within the DSLab environment, we work with language models with open weights, which can generate text, translate between languages, answer questions, summarise information and create source code. We initially deployed the Llama model, which we have gradually complemented with other advanced models such as Mistral and Granite. All models run entirely within the CNB’s internal environment, ensuring a high level of security when working with sensitive data.
Dynare 5.0 software platform for Matlab, intended for working with economic models. Dynare extends the analytical capabilities of DSLab and provides a tool for modelling, simulation and parameter estimation of economic systems in a unified, computation-ready environment.
Support for the Julia programming language. This extension has brought users new possibilities when working with economic models, both in terms of flexibility and computational speed.
- Julia is a modern programming language focused on scientific computing, known for its high speed and simple syntax, similar to Matlab or Python. It enables rapid prototyping as well as demanding numerical simulations, making it a suitable alternative to traditional tools. This allows for fast development and efficient execution of complex economic models, their simulations and other optimisation tasks.
- Integration into DSLab enables users to write and run Julia code without the need for local installation. The target group for the DSLab extension with Julia includes analysts, modellers and researchers engaged in macroeconomic modelling who require a powerful environment to develop and test their models.
Web-based IDE Visual Studio Code. This is the web version of the popular IDE (Integrated Development Environment) Visual Studio Code. Thanks to its extensibility (via installable extensions), IDE VS Code has become the de facto standard environment for developers and data analysts. The primary motivation for deploying this web-based version of VS Code in DSLab was to provide a more comfortable environment for developing and debugging Python code within DSLab.
In autumn 2025, we began testing the operation of large language models (LLMs) using an NVIDIA H100 graphics processing unit (GPU) on a loaned server. The aim is to verify whether this technology could be used to accelerate and streamline the operation of large language models already used internally by the CNB. Preliminary results indicate a significant increase in speed and output quality.