RBA develops internal AI chatbot

The chatbot is trained on 40 years of the central bank’s analytical documents.

The Reserve Bank of Australia has revealed a new internal AI-powered chatbot, RBAPubChat.

The Reserve Bank of Australia has developed and is testing an in-house, AI-powered chatbot to assist its staff with research and analysis.

Named RBAPubChat, the tool is trained on the central bank’s knowledge base of nearly 20,000 internal and external analytical documents spanning four decades. It aims to help employees ask policy-relevant questions and get useful summaries of existing information.

Speaking at the Shann memorial lecture in Perth, Governor Michele Bullock said that the AI is not being used to formulate or set monetary policy. Instead, it is intended to improve efficiency and amplify the impact of staff efforts.

A separate tool using natural language processing has also been developed to analyse over 22,000 conversations from the bank’s business liaison programme. The Reserve Bank of Australia has noted that this tool has already shown promise, helping to forecast wage growth more accurately than traditional models.

The RBA has also acquired its first enterprise-grade graphics processing unit to support developing and running advanced AI-driven tools.

The bank’s internal coding community is now a well-established part of its operations, with one in four employees using coding as a core part of their daily work. Governor Bullock stressed that the bank’s approach to technology is one of “deliberate, well-managed evolution” rather than disruption.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot