AI & Economics

Understanding the role of AI adoption

How is artificial intelligence reshaping public institutions and the economy? This page collects research, insights and reflections – on the macro-financial implications of AI, on its adoption inside policy institutions, and on the change management challenges this creates.

Insights

FEATUREDDecember 2025 · Working paper

Monetary Policy Tech – A primer to novel techniques for monetary policy analysis and business processes

with Douglas Araujo (Central Bank of Brazil; previously Bank for International Settlements)

We introduce "Monetary Policy Tech" – by analogy to SupTech and RegTech – to describe the rapidly growing use of AI and machine learning in central banking. The paper argues that AI affects central banks in two distinct ways: indirectly, through its impact on the macro-financial environment (productivity, market concentration, volatility, labour disruptions, inequality), and directly, through its impact on how central banks themselves do their work.

Diagram showing AI's impact on central banking: financial and real economy, central bank economic analyses, central bank business processes
Figure 1 – AI's impact on central banking, organised along three dimensions: the financial and real economy, central bank economic analyses, and central bank business processes.

We organise the direct impact along two axes. On the analysis side, we review how AI is transforming nowcasting (with alternative data such as satellite imagery, text and microdata), forecasting (where machine-learning models have beaten traditional benchmarks for inflation), macroeconomic modelling (deep learning to solve large DSGE/HANK models), and central bank communication (embeddings, sentiment analysis, even facial expressions of central bankers). On the business-process side, we cover document retrieval (chatbots and retrieval-augmented generation), data management (auto-cleaning, anomaly detection, web scraping) and document preparation (briefing tools, summarisation, consistency checks).

Throughout, one message remains core: humans stay the ultimate decisionmakers. The stochastic nature of AI outputs and the trade-offs between efficiency and confidentiality mean that governance and human oversight matter more, not less, as AI adoption deepens.

Key takeaways

  • AI hits central banks in two ways – through the economy they monitor, and through the work they do.
  • Machine learning now beats traditional benchmarks for inflation forecasting; alternative data (satellite, text, microdata) unlock new nowcasting signals.
  • Generative AI is moving from analysis into operations – document retrieval, briefing drafts, data commentary.
  • Governance, validation and human oversight are the binding constraints, not technology.
  • Three open questions: Does AI actually improve forecast accuracy? How do we validate AI for policy-grade work? Will AI agents on both sides of policy create new endogeneity loops?

Read the full paper on SSRN →

View all publications →