Research Discussion Paper – RDP 2025-06 An AI-powered Tool for Central Bank Business Liaisons: Quantitative Indicators and On-demand Insights from Firms

Contents

• Introduction
• RBA's Liaison Program
• Solution Architecture
• New Capabilities
• An Empirical Application: Nowcasting Wages Growth
• Conclusion
• Appendix A: Stratified Sampling to Assess Accuracy
• Appendix B: Spot-checking the Precision of the LM-based Topic Classifications
• Appendix C: Variable Description
• Appendix D: Variable Selection in the Lasso Model
• References

We would like to thank Markus Schwedeler, James Holloway and Claude Lopez for their helpful feedback and suggestions as well as seminar participants at the RBA/APRA Business Coding Knowledge Sharing Sessions and internal RBA seminars. There are also several people to thank for their contributions to the development of this tool. Max Zang, for his work on numerical extractions; Oliver Cutbill, for his work on the uncertainty index; Maureen Rocas, for helping to establish initial data pipelines; Joel Fernando, for his work on topic exposures; and Madison Terrell, for her patience and expertise helping us understand and source data to replicate current nowcasting processes at the RBA. We are also grateful to the entire Regional Industry and Analysis team (past and present) for their ongoing dedication to help us improve the tool and to all firms who participate in the RBA's liaison program. Finally, we are grateful to Paula Drew for her publications assistance. These are the views of the authors and not the Reserve Bank of Australia. Any remaining errors are our own.