Artificial Intelligence has the potential to have a greater impact on the global economy in the coming years and decades than almost any other technology before it. For central banks, AI is a great opportunity in their day-to-day work – but also an enormous challenge for their analyses. After all, it remains to be seen how the new technology will affect inflation, monetary policy, labour markets and economic growth.

Nevertheless, central banks need to take action now. The Bank for International Settlements (BIS) dedicated a special chapter to AI in its Annual Economic Report 2024, published at the end of June. The institution, also known as the central bank of central banks, called on central banks to make rapid progress in dealing with the technology. „Central banks need to come together and form a community to share knowledge, data, best practices and AI tools,“ it says. The technology could, for example, help central banks improve their forecasts for inflation and economic growth.

At the same time, however, the BIS also warns of the risks of AI for financial stability, for example, through new forms of cyber attacks. Furthermore, if many financial market participants were to make decisions on the basis of a handful of algorithms, this would increase the risk of bank runs and herd formation in general. It is, therefore, crucial for central banks to quickly understand the impact of AI on the economy and financial markets. „The use of AI is spreading faster than was the case with blockchain, for example,“ says Alexander Schroff, Financial Services Lead DACH at Publicis Sapient, which advises central banks on AI. The BIS also draws the same conclusion. AI has spread much faster in the USA than was the case in the early years of electricity, the internet, social media or smartphones.

According to Schroff, central banks have recognised the need for action, saying that „central banks are very active in the field of AI.“ ECB Director Piero Cipollone recently emphasised in a speech the importance of technology for monetary policy. New winners and losers in the capital and labour markets would have an impact on income and wealth distribution. „This is important for monetary policy, as it can influence people's propensity to consume and their access to credit, which in turn affects how demand reacts to changes in monetary policy“, he said. AI could also lead to an increase in non-bank-based financial intermediation. This would also have consequences for monetary policy transmission.

But what effect will AI have on inflation? There is no clear answer to this question. This makes it all the more important for central banks to keep a close eye on the issue. The BIS can imagine that inflation will initially be lower as a result of the technology, before the effect is reversed. In the short term, supply could exceed demand, which could reduce inflationary pressure. „But these effects could reverse over time if demand rises due to higher incomes,“ says the BIS. The central banks would have to take this dynamic into account.

Natascha Hinterlang, economist at the Bundesbank, also points out that AI can have different effects on price trends. „The influences on inflation are likely to be diverse and in some cases opposing. For example, price pressure decreases as productivity increases. Higher wage pressure due to rising demand for skilled labour and a strong additional demand for energy can work in the other direction,“ she says.

ECB Director Cipollone sees a further risk of inflation. „The use of AI will also have an impact on global energy demand, as the computing power required to maintain the rise of AI doubles every 100 days,“ he commented. „This could drive up energy costs.“

AI will increase the efficiency of companies – even if it is unclear how much. However, this does not necessarily lead to higher economic growth. Should the technology actually lead to an increase in hourly wages, employees could also decide to work less overall. Even apart from the discussion about the four-day week, there is a trend towards fewer working hours in many industrialised countries.

The effects on the labour market as a whole are also a major topic of discussion among economists. Which jobs will be lost? Which will change? Where will the technology even create new jobs? „AI has the potential to significantly increase productivity in some areas and change the labour market,“ says Hinterlang. „The exact macroeconomic and social influences are diverse and cannot yet be reliably estimated.“

Despite all the uncertainties that AI brings for the development of the economy, it could lead to better forecasts from central banks in the future, for example, by giving central banks a better insight into how prices and the economy will develop in the current month.

Based on a data set from the market research institute GfK, the Bundesbank has used machine learning techniques to produce an inflation forecast for the current month for a study. Central bankers refer to this as nowcasting. Nowcasting is of particular importance in dynamic phases.

GfK's data set includes daily household purchases for the period 2003 to 2022, with participants scanning the barcodes of the goods purchased at home. In its study, the Bundesbank concludes that analysing this data using machine learning provides a fairly accurate picture of inflation in the current month.

„We have identified four clusters where the use of AI can help us: For forecasts such as nowcasting, for the analysis of huge text-based data, for the use of complex, non-linear economic models and process optimisation,“ Felix Geiger, Deputy Head of the Capital Markets Department of the Bundesbank's Central Economics Division, comments with regard to the use of AI at the Bundesbank.

For example, the central bank could have speeches by ECB Governing Council members or the texts of monetary policy decisions analysed with AI to determine whether they come across as dovish or hawkish to the audience. „This enables us to better understand how and why the financial markets react to certain statements, for example,“ explains Geiger.

In June, the Bundesbank rolled out the „Text-based Intelligent Assistants“ (TIAs) platform. This enables employees of the central bank to summarise texts or have them written in their own style. This should significantly increase the efficiency of employees.

The application is based on the same language models as the famous ChatGPT. However, the models are not provided on American servers, but on French ones. This means that the Bundesbank's data cannot be used for further ChatGPT training. Data sovereignty remains with the Bundesbank. „Central banks must handle data very carefully when using AI so that the information does not fall into the hands of unauthorised persons,“ emphasises Schroff.

If this is taken into account, Artificial Intelligence can also make life much easier for central banks. „AI can help to identify the most relevant parts of large document databases in the shortest possible time,“ says Mauricio Fernández, Head of Technology at the Bundesbank's Strategy & Innovation department, listing another example of the sensible use of AI.

„The potential of AI is almost limitless,“ says Schroff. „One advantage of AI is that it is fundamentally not biased.“ Nevertheless, the results of AI can be. This is the case if the data records fed in have not been selected neutrally. It is, therefore, essential that the quality of the data that the AI works with is right. „Especially with quantitative data, I can imagine that AI can provide good advice. However, the decision ultimately lies with the human being.“