A BIS Bulletin Says a Focus on Retraining and Upskilling Central Bank Staff Is Warranted

On April 24, 2025, the Bank for International Settlements published a bulletin on what AI means for the people who work in central banks. BIS Bulletins are written by staff members of the BIS, and from time to time by other economists. This one says its views are those of its authors, and not necessarily the BIS's.

The paper says the path of AI is uncertain. So it sets out two scenarios, and asks what each would do to the staff. Under either one, it says a focus on retraining and upskilling existing staff, on attracting new talent and on a culture open to innovation "is warranted." Parts of the bulletin follow, not the whole.

Two Scenarios

The first scenario is the copilot. The paper defines a copilot based on a large language model as a tool that assists people with tasks such as developing software, summarizing documents, drafting email and generating images. The paper says central banks already use such tools in daily work. They make staff more productive "without fundamentally altering how their work is conducted."

The second scenario is the agent. An agent could replace people in "specific, well defined tasks with minimal human oversight." One example in the paper is an agent that gathers economic data in real time and keeps a forecast up to date. Another would automatically verify transactions conducted by central banks. The paper says such agents are still in beta and being tested at leading AI companies. Beta versions have shown mixed results in tasks such as booking a flight.

Even here, the paper says, human oversight "will remain crucial." A "human in the loop" is still needed to interpret findings and make high-level decisions. In this scenario, some tasks of central bank staff could be done by AI. That could reduce the need for some roles, the paper says, while creating new tasks at the same time.

What Changes for the Staff

In its introduction, the paper says tasks such as real-time economic forecasting, financial stability monitoring and verifying transactions will increasingly rely on AI. That will require staff to oversee, interpret and refine what the AI produces. In either scenario, it says, job profiles will change. New roles will be needed to build and maintain AI tools. Existing roles will need added skills. Some roles built on more routine or repetitive tasks "may become less important."

With copilots, the paper says, the focus will be on upskilling and reskilling. That training would need to be continuous, because the tools improve so fast. Experts in statistics, law, finance or economics, and professionals in support functions like IT, HR and security, will need to work AI tools into their tasks and interpret what the tools produce.

With agents, the focus of the workforce strategy would be on managing the transition. Tasks would be redefined to put "a greater emphasis on oversight and supervision." Staff who are not AI specialists would need to understand how the autonomous tools work, and what they can and cannot do.

The training differs too. For copilots, the paper says, training will focus on collaboration skills and on integrating AI insights. For agents, the emphasis will be on advanced technical skills, system management and supervision of AI tools.

Hiring and Skills

The paper draws on a BIS survey of 52 members of the Central Bank Governance Network, taken in May 2024. Almost 90% of the central banks surveyed said recruiting had become more difficult in the last five years. Hiring was particularly difficult in areas such as cyber security, IT, fintech, data science and AI. One reason, the paper says, is high demand for these roles across the economy. Public institutions may not be able to match private sector salaries for top AI talent.

The paper also reports that 83% of central banks reported more complexity in workforce planning.

Some central banks face legal or regulatory limits on recruitment. In the survey, 58% of respondents faced them. Sourcing requirements, such as hiring only through civil service exams, were cited by 47%. Citizenship requirements were cited by 43%.

There is also the staff already inside. Many central banks, the paper says, were concerned about whether experienced staff could acquire new skills and ways of working. Several noted challenges in acquiring enough digital, AI or advanced data analytics skills at the needed pace. That may hinder the adoption of new tools.

Retraining and Culture

The paper says internal and external experts will be needed to retrain staff to use AI tools. Central banks themselves, it says, have stressed a culture of greater innovation. They also recognize it may come at a more cautious pace than in the private sector, "given their lower risk appetites and need for broad stakeholder engagement."